PowerNV: make PowerNV PCIe working on a real hardware

Fixes:
- map all devices to PE0
- use 1:1 TCE mapping
- provide the same TCE mapping for all PEs (not only PE0)
- add TCE reset and alignment (required by OPAL)

Created by:            Wojciech Macek <wma@semihalf.com>
Obtained from:         Semihalf
Sponsored by:          QCM Technologies
This commit is contained in:
Wojciech Macek 2018-01-17 07:39:11 +00:00
parent 957fc241ec
commit 5c3e53ef19
2 changed files with 81 additions and 47 deletions

View File

@ -65,10 +65,13 @@ int opal_call(uint64_t token, ...);
#define OPAL_GET_MSI_32 39
#define OPAL_GET_MSI_64 40
#define OPAL_PCI_MSI_EOI 63
#define OPAL_PCI_GET_PHB_DIAG_DATA2 64
#define OPAL_START_CPU 41
#define OPAL_PCI_MAP_PE_DMA_WINDOW 44
#define OPAL_PCI_MAP_PE_DMA_WINDOW_REAL 45
#define OPAL_RETURN_CPU 69
#define OPAL_REINIT_CPUS 70
#define OPAL_PCI_TCE_KILL 126
/* For OPAL_PCI_SET_PE */
#define OPAL_UNMAP_PE 0

View File

@ -1,5 +1,6 @@
/*-
* Copyright (c) 2015-2016 Nathan Whitehorn
* Copyright (c) 2017-2018 Semihalf
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
@ -58,6 +59,12 @@ __FBSDID("$FreeBSD$");
#include "iommu_if.h"
#include "opal.h"
#define OPAL_PCI_TCE_MAX_ENTRIES (1024*1024UL)
#define OPAL_PCI_TCE_SEG_SIZE (16*1024*1024UL)
#define OPAL_PCI_TCE_R (1UL << 0)
#define OPAL_PCI_TCE_W (1UL << 1)
#define PHB3_TCE_KILL_INVAL_ALL (1UL << 63)
/*
* Device interface.
*/
@ -148,6 +155,8 @@ struct opalpci_softc {
vmem_t *msi_vmem;
int msi_base; /* Base XIVE number */
int base_msi_irq; /* Base IRQ assigned by FreeBSD to this PIC */
uint64_t *tce; /* TCE table for 1:1 mapping */
struct resource *r_reg;
};
static devclass_t opalpci_devclass;
@ -177,12 +186,24 @@ opalpci_probe(device_t dev)
return (BUS_PROBE_GENERIC);
}
static void
pci_phb3_tce_invalidate_entire(struct opalpci_softc *sc)
{
mb();
bus_write_8(sc->r_reg, 0x210, PHB3_TCE_KILL_INVAL_ALL);
mb();
}
static int
opalpci_attach(device_t dev)
{
struct opalpci_softc *sc;
cell_t id[2], m64window[6], npe;
int i, err;
uint64_t maxmem;
uint64_t entries;
int rid;
sc = device_get_softc(dev);
@ -204,6 +225,15 @@ opalpci_attach(device_t dev)
if (bootverbose)
device_printf(dev, "OPAL ID %#lx\n", sc->phb_id);
rid = 0;
sc->r_reg = bus_alloc_resource_any(dev, SYS_RES_MEMORY,
&rid, RF_ACTIVE | RF_SHAREABLE);
if (sc->r_reg == NULL) {
device_printf(dev, "Failed to allocate PHB[%jd] registers\n",
(uintmax_t)sc->phb_id);
return (ENXIO);
}
/*
* Reset PCI IODA table
*/
@ -214,46 +244,11 @@ opalpci_attach(device_t dev)
return (ENXIO);
}
while ((err = opal_call(OPAL_PCI_POLL, sc->phb_id)) > 0)
DELAY(1000*err); /* Returns expected delay in ms */
DELAY(1000*(err + 1)); /* Returns expected delay in ms */
if (err < 0) {
device_printf(dev, "PHB IODA reset poll failed: %d\n", err);
return (ENXIO);
device_printf(dev, "WARNING: PHB IODA reset poll failed: %d\n", err);
}
/*
* Reset everything. Especially important if we have inherited the
* system from Linux by kexec()
*/
#ifdef NOTYET
if (bootverbose)
device_printf(dev, "Resetting PCI bus\n");
err = opal_call(OPAL_PCI_RESET, sc->phb_id, OPAL_RESET_PHB_COMPLETE, 1);
if (err < 0) {
device_printf(dev, "PHB reset failed: %d\n", err);
return (ENXIO);
}
while ((err = opal_call(OPAL_PCI_POLL, sc->phb_id)) > 0)
DELAY(1000*err); /* Returns expected delay in ms */
if (err < 0) {
device_printf(dev, "PHB reset poll failed: %d\n", err);
return (ENXIO);
}
DELAY(10000);
err = opal_call(OPAL_PCI_RESET, sc->phb_id, OPAL_RESET_PHB_COMPLETE, 0);
if (err < 0) {
device_printf(dev, "PHB reset completion failed: %d\n", err);
return (ENXIO);
}
while ((err = opal_call(OPAL_PCI_POLL, sc->phb_id)) > 0)
DELAY(1000*err); /* Returns expected delay in ms */
if (err < 0) {
device_printf(dev, "PHB reset completion poll failed: %d\n",
err);
return (ENXIO);
}
DELAY(10000);
#endif
/*
* Map all devices on the bus to partitionable endpoint one until
* such time as we start wanting to do things like bhyve.
@ -282,6 +277,8 @@ opalpci_attach(device_t dev)
/* XXX: multiple M64 windows? */
if (OF_getencprop(ofw_bus_get_node(dev), "ibm,opal-m64-window",
m64window, sizeof(m64window)) == sizeof(m64window)) {
opal_call(OPAL_PCI_PHB_MMIO_ENABLE, sc->phb_id,
OPAL_M64_WINDOW_TYPE, 0, 0);
opal_call(OPAL_PCI_SET_PHB_MEM_WINDOW, sc->phb_id,
OPAL_M64_WINDOW_TYPE, 0 /* index */,
((uint64_t)m64window[2] << 32) | m64window[3], 0,
@ -294,20 +291,53 @@ opalpci_attach(device_t dev)
}
/*
* Also disable the IOMMU for the time being for PE 1 (everything)
* Enable IOMMU for PE1 - map everything 1:1 using
* segments of OPAL_PCI_TCE_SEG_SIZE size
*/
maxmem = roundup2(powerpc_ptob(Maxmem), OPAL_PCI_TCE_SEG_SIZE);
entries = maxmem / OPAL_PCI_TCE_SEG_SIZE;
if (entries > OPAL_PCI_TCE_MAX_ENTRIES)
panic("POWERNV supports only %jdGB of memory space\n",
(uintmax_t)((OPAL_PCI_TCE_MAX_ENTRIES * OPAL_PCI_TCE_SEG_SIZE) >> 30));
if (bootverbose)
device_printf(dev, "Mapping 0-%#lx for DMA\n",
roundup2(powerpc_ptob(Maxmem), 16*1024*1024));
err = opal_call(OPAL_PCI_MAP_PE_DMA_WINDOW_REAL, sc->phb_id,
OPAL_PCI_DEFAULT_PE, OPAL_PCI_DEFAULT_PE << 1,
0 /* start address */, roundup2(powerpc_ptob(Maxmem),
16*1024*1024)/* all RAM */);
if (err != 0) {
device_printf(dev, "DMA mapping failed: %d\n", err);
return (ENXIO);
device_printf(dev, "Mapping 0-%#jx for DMA\n", (uintmax_t)maxmem);
sc->tce = contigmalloc(OPAL_PCI_TCE_MAX_ENTRIES * sizeof(uint64_t),
M_DEVBUF, M_NOWAIT | M_ZERO, 0,
BUS_SPACE_MAXADDR_32BIT, OPAL_PCI_TCE_SEG_SIZE, 0);
if (sc->tce == NULL)
panic("Failed to allocate TCE memory for PHB %jd\n",
(uintmax_t)sc->phb_id);
for (i = 0; i < entries; i++)
sc->tce[i] = (i * OPAL_PCI_TCE_SEG_SIZE) | OPAL_PCI_TCE_R | OPAL_PCI_TCE_W;
/* Map TCE for every PE. It seems necessary for Power8 */
for (i = 0; i < npe; i++) {
err = opal_call(OPAL_PCI_MAP_PE_DMA_WINDOW, sc->phb_id,
i, (i << 1),
1, pmap_kextract((uint64_t)&sc->tce[0]),
OPAL_PCI_TCE_MAX_ENTRIES * sizeof(uint64_t), OPAL_PCI_TCE_SEG_SIZE);
if (err != 0) {
device_printf(dev, "DMA IOMMU mapping failed: %d\n", err);
return (ENXIO);
}
err = opal_call(OPAL_PCI_MAP_PE_DMA_WINDOW_REAL, sc->phb_id,
i, (i << 1) + 1,
(1UL << 59), maxmem);
if (err != 0) {
device_printf(dev, "DMA 64b bypass mapping failed: %d\n", err);
return (ENXIO);
}
}
/*
* Invalidate all previous TCE entries.
*
* TODO: add support for other PHBs than PHB3
*/
pci_phb3_tce_invalidate_entire(sc);
/*
* Get MSI properties
*/
@ -457,6 +487,7 @@ opalpci_write_config(device_t dev, u_int bus, u_int slot, u_int func,
static int
opalpci_route_interrupt(device_t bus, device_t dev, int pin)
{
return (pin);
}