Hi! My word...
> What is your guidance, should we introduce new memory maps for GICv3
> enabled machine or should we move to a single dynamic memory map?
IMHO there's no reason to introduce another memory map. I have already done test integration some time ago, and here is what i got:
--- cut ---
enum {
VIRT_FLASH,
VIRT_MEM,
VIRT_CPUPERIPHS,
VIRT_GIC_DIST,
VIRT_GIC_CPU,
VIRT_GIC_DIST_SPI = VIRT_GIC_CPU,
VIRT_ITS_CONTROL,
VIRT_ITS_TRANSLATION,
VIRT_LPI,
VIRT_UART,
VIRT_MMIO,
VIRT_RTC,
VIRT_FW_CFG,
VIRT_PCIE,
VIRT_GIC_V2M = VIRT_ITS_CONTROL,
};
--- cut ---
static const MemMapEntry a15memmap[] = {
/* Space up to 0x8000000 is reserved for a boot ROM */
[VIRT_FLASH] = { 0, 0x08000000 },
[VIRT_CPUPERIPHS] = { 0x08000000, 0x00020000 },
/* GIC distributor and CPU interfaces sit inside the CPU peripheral space */
[VIRT_GIC_DIST] = { 0x08000000, 0x00010000 },
[VIRT_GIC_CPU] = { 0x08010000, 0x00010000 }, /* VIRT_GIC_DIST_SPI for v3 */
[VIRT_ITS_CONTROL] = { 0x08020000, 0x0010000 }, /* VIRT_GIC_V2M for v2 */
[VIRT_ITS_TRANSLATION] = { 0x08030000, 0x00010000 },
[VIRT_LPI] = { 0x08040000, 0x00800000 },
[VIRT_UART] = { 0x09000000, 0x00001000 },
[VIRT_RTC] = { 0x09010000, 0x00001000 },
[VIRT_FW_CFG] = { 0x09020000, 0x0000000a },
[VIRT_MMIO] = { 0x0a000000, 0x00000200 },
/* ...repeating for a total of NUM_VIRTIO_TRANSPORTS, each of that size */
/*
* PCIE verbose map:
*
* MMIO window { 0x10000000, 0x2eff0000 },
* PIO window { 0x3eff0000, 0x00010000 },
* ECAM { 0x3f000000, 0x01000000 },
*/
[VIRT_PCIE] = { 0x10000000, 0x30000000 },
[VIRT_MEM] = { 0x40000000, 30ULL * 1024 * 1024 * 1024 },
};
--- cut ---
As you can see, it's perfectly readable and memory maps are identical. I even see no problem with merging ITS_CONTROL and ITS_TRANSPATION, because i don't see any harm in increasing v2m area.
P.S. And yes, VIRT_GIC_DIST_SPI should be VIRT_GIC_DIST_MBI instead (Reviewed-by: Eric Auger <eric.auger@xxxxxxxxxx>), just this fragment is from my old integration branch, which i currently don't work on, because my new test environment doesn't use GICv2 at all.
Kind regards,
Pavel Fedin
Expert Engineer
Samsung Electronics Research center Russia
_______________________________________________
kvmarm mailing list
kvmarm@xxxxxxxxxxxxxxxxxxxxx
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
