[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH qemu v18 5/5] spapr_pci/spapr_pci_vfio: Support
From: |
Alexey Kardashevskiy |
Subject: |
Re: [Qemu-devel] [PATCH qemu v18 5/5] spapr_pci/spapr_pci_vfio: Support Dynamic DMA Windows (DDW) |
Date: |
Wed, 22 Jun 2016 18:26:52 +1000 |
User-agent: |
Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Thunderbird/45.1.1 |
On 22/06/16 17:01, David Gibson wrote:
> On Wed, Jun 22, 2016 at 01:23:51PM +1000, Alexey Kardashevskiy wrote:
>> On 22/06/16 12:35, David Gibson wrote:
>>> On Tue, Jun 21, 2016 at 11:14:05AM +1000, Alexey Kardashevskiy wrote:
>>>> This adds support for Dynamic DMA Windows (DDW) option defined by
>>>> the SPAPR specification which allows to have additional DMA window(s)
>>>>
>>>> The "ddw" property is enabled by default on a PHB but for compatibility
>>>> the pseries-2.6 machine and older disable it.
>>>> This also creates a single DMA window for the older machines to
>>>> maintain backward migration.
>>>>
>>>> This implements DDW for PHB with emulated and VFIO devices. The host
>>>> kernel support is required. The advertised IOMMU page sizes are 4K and
>>>> 64K; 16M pages are supported but not advertised by default, in order to
>>>> enable them, the user has to specify "pgsz" property for PHB and
>>>> enable huge pages for RAM.
>>>>
>>>> The existing linux guests try creating one additional huge DMA window
>>>> with 64K or 16MB pages and map the entire guest RAM to. If succeeded,
>>>> the guest switches to dma_direct_ops and never calls TCE hypercalls
>>>> (H_PUT_TCE,...) again. This enables VFIO devices to use the entire RAM
>>>> and not waste time on map/unmap later. This adds a "dma64_win_addr"
>>>> property which is a bus address for the 64bit window and by default
>>>> set to 0x800.0000.0000.0000 as this is what the modern POWER8 hardware
>>>> uses and this allows having emulated and VFIO devices on the same bus.
>>>>
>>>> This adds 4 RTAS handlers:
>>>> * ibm,query-pe-dma-window
>>>> * ibm,create-pe-dma-window
>>>> * ibm,remove-pe-dma-window
>>>> * ibm,reset-pe-dma-window
>>>> These are registered from type_init() callback.
>>>>
>>>> These RTAS handlers are implemented in a separate file to avoid polluting
>>>> spapr_iommu.c with PCI.
>>>>
>>>> This changes sPAPRPHBState::dma_liobn to an array to allow 2 LIOBNs.
>>>>
>>>> Signed-off-by: Alexey Kardashevskiy <address@hidden>
>>>
>>> A few queries below. Not sure if they'll require code changes or just
>>> explanation.
>>>
>>>> ---
>>>> Changes:
>>>> v18:
>>>> * fixed bug when ddw-create rtas call was always creating window at 1<<59
>>>> offset
>>>> * update minimum supported machine version
>>>> * s/dma64_window_addr/dma_win_addr/ to match dma_win_addr
>>>>
>>>> v17:
>>>> * fixed: "query" did return non-page-shifted value when memory hotplug is
>>>> enabled
>>>>
>>>> v16:
>>>> * s/dma_liobn/dma_liobn[SPAPR_PCI_DMA_MAX_WINDOWS]/
>>>> * s/SPAPR_PCI_LIOBN()/dma_liobn[]/
>>>>
>>>> v15:
>>>> * moved page mask filtering to PHB realize(), use "-mempath" to know
>>>> if there are huge pages
>>>> * fixed error reporting in RTAS handlers
>>>> * max window size accounts now hotpluggable memory boundaries
>>>> ---
>>>> hw/ppc/Makefile.objs | 1 +
>>>> hw/ppc/spapr.c | 7 +-
>>>> hw/ppc/spapr_pci.c | 77 +++++++++---
>>>> hw/ppc/spapr_rtas_ddw.c | 295
>>>> ++++++++++++++++++++++++++++++++++++++++++++
>>>> include/hw/pci-host/spapr.h | 8 +-
>>>> include/hw/ppc/spapr.h | 16 ++-
>>>> trace-events | 4 +
>>>> 7 files changed, 386 insertions(+), 22 deletions(-)
>>>> create mode 100644 hw/ppc/spapr_rtas_ddw.c
>>>>
>>>> diff --git a/hw/ppc/Makefile.objs b/hw/ppc/Makefile.objs
>>>> index 5cc6608..91a3420 100644
>>>> --- a/hw/ppc/Makefile.objs
>>>> +++ b/hw/ppc/Makefile.objs
>>>> @@ -8,6 +8,7 @@ obj-$(CONFIG_PSERIES) += spapr_cpu_core.o
>>>> ifeq ($(CONFIG_PCI)$(CONFIG_PSERIES)$(CONFIG_LINUX), yyy)
>>>> obj-y += spapr_pci_vfio.o
>>>> endif
>>>> +obj-$(CONFIG_PSERIES) += spapr_rtas_ddw.o
>>>> # PowerPC 4xx boards
>>>> obj-y += ppc405_boards.o ppc4xx_devs.o ppc405_uc.o ppc440_bamboo.o
>>>> obj-y += ppc4xx_pci.o
>>>> diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
>>>> index 778fa25..f7cff27 100644
>>>> --- a/hw/ppc/spapr.c
>>>> +++ b/hw/ppc/spapr.c
>>>> @@ -2485,7 +2485,12 @@ DEFINE_SPAPR_MACHINE(2_7, "2.7", true);
>>>> * pseries-2.6
>>>> */
>>>> #define SPAPR_COMPAT_2_6 \
>>>> - HW_COMPAT_2_6
>>>> + HW_COMPAT_2_6 \
>>>> + { \
>>>> + .driver = TYPE_SPAPR_PCI_HOST_BRIDGE,\
>>>> + .property = "ddw",\
>>>> + .value = stringify(off),\
>>>> + },
>>>>
>>>> static void spapr_machine_2_6_instance_options(MachineState *machine)
>>>> {
>>>> diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
>>>> index 9f28fb3..0cb51dd 100644
>>>> --- a/hw/ppc/spapr_pci.c
>>>> +++ b/hw/ppc/spapr_pci.c
>>>> @@ -35,6 +35,7 @@
>>>> #include "hw/ppc/spapr.h"
>>>> #include "hw/pci-host/spapr.h"
>>>> #include "exec/address-spaces.h"
>>>> +#include "exec/ram_addr.h"
>>>> #include <libfdt.h>
>>>> #include "trace.h"
>>>> #include "qemu/error-report.h"
>>>> @@ -45,6 +46,7 @@
>>>> #include "hw/ppc/spapr_drc.h"
>>>> #include "sysemu/device_tree.h"
>>>> #include "sysemu/kvm.h"
>>>> +#include "sysemu/hostmem.h"
>>>>
>>>> #include "hw/vfio/vfio.h"
>>>>
>>>> @@ -1088,7 +1090,7 @@ static void
>>>> spapr_phb_add_pci_device(sPAPRDRConnector *drc,
>>>> int fdt_start_offset = 0, fdt_size;
>>>>
>>>> if (object_dynamic_cast(OBJECT(pdev), "vfio-pci")) {
>>>> - sPAPRTCETable *tcet = spapr_tce_find_by_liobn(phb->dma_liobn);
>>>> + sPAPRTCETable *tcet = spapr_tce_find_by_liobn(phb->dma_liobn[0]);
>>>>
>>>> spapr_tce_set_need_vfio(tcet, true);
>>>
>>> Now that Alex took your notifier on/off patches, can you remove this
>>> chunk?
>>
>> It will stop compiling as dma_liobn is an array now.
>
> Sorry, I wasn't clear. I meant remove this whole if statement, not
> just remove this hunk of the patch.
Bisect-ability will suffer then, and we can easily avoided if this patch is
applied on top of these:
vfio, memory: Notify IOMMU about starting/stopping listening
spapr_iommu: Realloc guest visible TCE table when starting/stopping listening
All we need is Alex to send pull req, Peter to merge it and you to rebase
ppc-for-2.7 on top of this :)
>
>>> If it's still necessary, don't you need to loop over all the
>>> possible liobns, rather than just acting on liobn[0]?
>>
>> Ah, right. Forgot about it. That was the reason why I wanted those notifier
>> callbacks in this series, lost it in respins. I do need a loop here which
>> I'll have to remove soon though.
>
> Ok.
>>>> @@ -1527,6 +1544,11 @@ static Property spapr_phb_properties[] = {
>>>> /* Default DMA window is 0..1GB */
>>>> DEFINE_PROP_UINT64("dma_win_addr", sPAPRPHBState, dma_win_addr, 0),
>>>> DEFINE_PROP_UINT64("dma_win_size", sPAPRPHBState, dma_win_size,
>>>> 0x40000000),
>>>> + DEFINE_PROP_UINT64("dma64_win_addr", sPAPRPHBState, dma64_win_addr,
>>>> + 0x800000000000000ULL),
>>>> + DEFINE_PROP_BOOL("ddw", sPAPRPHBState, ddw_enabled, true),
>>>> + DEFINE_PROP_UINT64("pgsz", sPAPRPHBState, page_size_mask,
>>>> + (1ULL << 12) | (1ULL << 16)),
>>>> DEFINE_PROP_END_OF_LIST(),
>>>> };
>>>>
>>>> @@ -1603,7 +1625,7 @@ static const VMStateDescription vmstate_spapr_pci = {
>>>> .post_load = spapr_pci_post_load,
>>>> .fields = (VMStateField[]) {
>>>> VMSTATE_UINT64_EQUAL(buid, sPAPRPHBState),
>>>> - VMSTATE_UINT32_EQUAL(dma_liobn, sPAPRPHBState),
>>>> + VMSTATE_UNUSED(4), /* dma_liobn */
>>>
>>> It's not obvious to me why this change is necessary.
>>
>> It is not. But I was touching liobn and this is a proper cleanup which
>> needs to be done anyway as _EQUAL() macros are sort of deprecated and
>> rather pointless. Since I am adding a new 64bit LIOBN in this patch, should
>> I add it in VMSTATE as 32bit one and bump the vmstate version? Or not add
>> it (leaving some inconsistency)?
>
> Ah, ok, I see your point. Yeah, I guess we can drop it.
>
>>>> VMSTATE_UINT64_EQUAL(mem_win_addr, sPAPRPHBState),
>>>> VMSTATE_UINT64_EQUAL(mem_win_size, sPAPRPHBState),
>>>> VMSTATE_UINT64_EQUAL(io_win_addr, sPAPRPHBState),
>>>> @@ -1779,6 +1801,15 @@ int spapr_populate_pci_dt(sPAPRPHBState *phb,
>>>> uint32_t interrupt_map_mask[] = {
>>>> cpu_to_be32(b_ddddd(-1)|b_fff(0)), 0x0, 0x0, cpu_to_be32(-1)};
>>>> uint32_t interrupt_map[PCI_SLOT_MAX * PCI_NUM_PINS][7];
>>>> + uint32_t ddw_applicable[] = {
>>>> + cpu_to_be32(RTAS_IBM_QUERY_PE_DMA_WINDOW),
>>>> + cpu_to_be32(RTAS_IBM_CREATE_PE_DMA_WINDOW),
>>>> + cpu_to_be32(RTAS_IBM_REMOVE_PE_DMA_WINDOW)
>>>> + };
>>>> + uint32_t ddw_extensions[] = {
>>>> + cpu_to_be32(1),
>>>> + cpu_to_be32(RTAS_IBM_RESET_PE_DMA_WINDOW)
>>>> + };
>>>> sPAPRTCETable *tcet;
>>>> PCIBus *bus = PCI_HOST_BRIDGE(phb)->bus;
>>>> sPAPRFDT s_fdt;
>>>> @@ -1803,6 +1834,14 @@ int spapr_populate_pci_dt(sPAPRPHBState *phb,
>>>> _FDT(fdt_setprop_cell(fdt, bus_off, "ibm,pci-config-space-type",
>>>> 0x1));
>>>> _FDT(fdt_setprop_cell(fdt, bus_off, "ibm,pe-total-#msi", XICS_IRQS));
>>>>
>>>> + /* Dynamic DMA window */
>>>> + if (phb->ddw_enabled) {
>>>> + _FDT(fdt_setprop(fdt, bus_off, "ibm,ddw-applicable",
>>>> &ddw_applicable,
>>>> + sizeof(ddw_applicable)));
>>>> + _FDT(fdt_setprop(fdt, bus_off, "ibm,ddw-extensions",
>>>> + &ddw_extensions, sizeof(ddw_extensions)));
>>>> + }
>>>> +
>>>> /* Build the interrupt-map, this must matches what is done
>>>> * in pci_spapr_map_irq
>>>> */
>>>> @@ -1826,7 +1865,7 @@ int spapr_populate_pci_dt(sPAPRPHBState *phb,
>>>> _FDT(fdt_setprop(fdt, bus_off, "interrupt-map", &interrupt_map,
>>>> sizeof(interrupt_map)));
>>>>
>>>> - tcet = spapr_tce_find_by_liobn(phb->dma_liobn);
>>>> + tcet = spapr_tce_find_by_liobn(phb->dma_liobn[0]);
>>>> if (!tcet) {
>>>> return -1;
>>>> }
>>>> diff --git a/hw/ppc/spapr_rtas_ddw.c b/hw/ppc/spapr_rtas_ddw.c
>>>> new file mode 100644
>>>> index 0000000..177dcff
>>>> --- /dev/null
>>>> +++ b/hw/ppc/spapr_rtas_ddw.c
>>>> @@ -0,0 +1,295 @@
>>>> +/*
>>>> + * QEMU sPAPR Dynamic DMA windows support
>>>> + *
>>>> + * Copyright (c) 2015 Alexey Kardashevskiy, IBM Corporation.
>>>> + *
>>>> + * This program is free software; you can redistribute it and/or modify
>>>> + * it under the terms of the GNU General Public License as published by
>>>> + * the Free Software Foundation; either version 2 of the License,
>>>> + * or (at your option) any later version.
>>>> + *
>>>> + * This program is distributed in the hope that it will be useful,
>>>> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
>>>> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
>>>> + * GNU General Public License for more details.
>>>> + *
>>>> + * You should have received a copy of the GNU General Public License
>>>> + * along with this program; if not, see <http://www.gnu.org/licenses/>.
>>>> + */
>>>> +
>>>> +#include "qemu/osdep.h"
>>>> +#include "cpu.h"
>>>> +#include "qemu/error-report.h"
>>>> +#include "hw/ppc/spapr.h"
>>>> +#include "hw/pci-host/spapr.h"
>>>> +#include "trace.h"
>>>> +
>>>> +static int spapr_phb_get_active_win_num_cb(Object *child, void *opaque)
>>>> +{
>>>> + sPAPRTCETable *tcet;
>>>> +
>>>> + tcet = (sPAPRTCETable *) object_dynamic_cast(child,
>>>> TYPE_SPAPR_TCE_TABLE);
>>>> + if (tcet && tcet->nb_table) {
>>>> + ++*(unsigned *)opaque;
>>>> + }
>>>> + return 0;
>>>> +}
>>>> +
>>>> +static unsigned spapr_phb_get_active_win_num(sPAPRPHBState *sphb)
>>>> +{
>>>> + unsigned ret = 0;
>>>> +
>>>> + object_child_foreach(OBJECT(sphb), spapr_phb_get_active_win_num_cb,
>>>> &ret);
>>>> +
>>>> + return ret;
>>>> +}
>>>> +
>>>> +static int spapr_phb_get_free_liobn_cb(Object *child, void *opaque)
>>>> +{
>>>> + sPAPRTCETable *tcet;
>>>> +
>>>> + tcet = (sPAPRTCETable *) object_dynamic_cast(child,
>>>> TYPE_SPAPR_TCE_TABLE);
>>>> + if (tcet && !tcet->nb_table) {
>>>> + *(uint32_t *)opaque = tcet->liobn;
>>>> + return 1;
>>>> + }
>>>> + return 0;
>>>> +}
>>>> +
>>>> +static unsigned spapr_phb_get_free_liobn(sPAPRPHBState *sphb)
>>>> +{
>>>> + uint32_t liobn = 0;
>>>> +
>>>> + object_child_foreach(OBJECT(sphb), spapr_phb_get_free_liobn_cb,
>>>> &liobn);
>>>> +
>>>> + return liobn;
>>>> +}
>>>> +
>>>> +static uint32_t spapr_page_mask_to_query_mask(uint64_t page_mask)
>>>> +{
>>>> + int i;
>>>> + uint32_t mask = 0;
>>>> + const struct { int shift; uint32_t mask; } masks[] = {
>>>> + { 12, RTAS_DDW_PGSIZE_4K },
>>>> + { 16, RTAS_DDW_PGSIZE_64K },
>>>> + { 24, RTAS_DDW_PGSIZE_16M },
>>>> + { 25, RTAS_DDW_PGSIZE_32M },
>>>> + { 26, RTAS_DDW_PGSIZE_64M },
>>>> + { 27, RTAS_DDW_PGSIZE_128M },
>>>> + { 28, RTAS_DDW_PGSIZE_256M },
>>>> + { 34, RTAS_DDW_PGSIZE_16G },
>>>> + };
>>>> +
>>>> + for (i = 0; i < ARRAY_SIZE(masks); ++i) {
>>>> + if (page_mask & (1ULL << masks[i].shift)) {
>>>> + mask |= masks[i].mask;
>>>> + }
>>>> + }
>>>> +
>>>> + return mask;
>>>> +}
>>>> +
>>>> +static void rtas_ibm_query_pe_dma_window(PowerPCCPU *cpu,
>>>> + sPAPRMachineState *spapr,
>>>> + uint32_t token, uint32_t nargs,
>>>> + target_ulong args,
>>>> + uint32_t nret, target_ulong rets)
>>>> +{
>>>> + sPAPRPHBState *sphb;
>>>> + uint64_t buid, max_window_size;
>>>> + uint32_t avail, addr, pgmask = 0;
>>>> + MachineState *machine = MACHINE(spapr);
>>>> +
>>>> + if ((nargs != 3) || (nret != 5)) {
>>>> + goto param_error_exit;
>>>> + }
>>>> +
>>>> + buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
>>>> + addr = rtas_ld(args, 0);
>>>> + sphb = spapr_pci_find_phb(spapr, buid);
>>>> + if (!sphb || !sphb->ddw_enabled) {
>>>> + goto param_error_exit;
>>>> + }
>>>> +
>>>> + /* Translate page mask to LoPAPR format */
>>>> + pgmask = spapr_page_mask_to_query_mask(sphb->page_size_mask);
>>>> +
>>>> + /*
>>>> + * This is "Largest contiguous block of TCEs allocated specifically
>>>> + * for (that is, are reserved for) this PE".
>>>> + * Return the maximum number as maximum supported RAM size was in 4K
>>>> pages.
>>>> + */
>>>> + if (machine->ram_size == machine->maxram_size) {
>>>> + max_window_size = machine->ram_size;
>>>> + } else {
>>>> + MemoryHotplugState *hpms = &spapr->hotplug_memory;
>>>> +
>>>> + max_window_size = hpms->base + memory_region_size(&hpms->mr);
>>>> + }
>>>> +
>>>> + avail = SPAPR_PCI_DMA_MAX_WINDOWS -
>>>> spapr_phb_get_active_win_num(sphb);
>>>> +
>>>> + rtas_st(rets, 0, RTAS_OUT_SUCCESS);
>>>> + rtas_st(rets, 1, avail);
>>>> + rtas_st(rets, 2, max_window_size >> SPAPR_TCE_PAGE_SHIFT);
>>>> + rtas_st(rets, 3, pgmask);
>>>> + rtas_st(rets, 4, 0); /* DMA migration mask, not supported */
>>>> +
>>>> + trace_spapr_iommu_ddw_query(buid, addr, avail, max_window_size,
>>>> pgmask);
>>>> + return;
>>>> +
>>>> +param_error_exit:
>>>> + rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
>>>> +}
>>>> +
>>>> +static void rtas_ibm_create_pe_dma_window(PowerPCCPU *cpu,
>>>> + sPAPRMachineState *spapr,
>>>> + uint32_t token, uint32_t nargs,
>>>> + target_ulong args,
>>>> + uint32_t nret, target_ulong
>>>> rets)
>>>> +{
>>>> + sPAPRPHBState *sphb;
>>>> + sPAPRTCETable *tcet = NULL;
>>>> + uint32_t addr, page_shift, window_shift, liobn;
>>>> + uint64_t buid, win_addr;
>>>> + int windows;
>>>> +
>>>> + if ((nargs != 5) || (nret != 4)) {
>>>> + goto param_error_exit;
>>>> + }
>>>> +
>>>> + buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
>>>> + addr = rtas_ld(args, 0);
>>>> + sphb = spapr_pci_find_phb(spapr, buid);
>>>> + if (!sphb || !sphb->ddw_enabled) {
>>>> + goto param_error_exit;
>>>> + }
>>>> +
>>>> + page_shift = rtas_ld(args, 3);
>>>> + window_shift = rtas_ld(args, 4);
>>>> + liobn = spapr_phb_get_free_liobn(sphb);
>>>> + windows = spapr_phb_get_active_win_num(sphb);
>>>> +
>>>> + if (!(sphb->page_size_mask & (1ULL << page_shift)) ||
>>>> + (window_shift < page_shift)) {
>>>> + goto param_error_exit;
>>>> + }
>>>> +
>>>> + if (!liobn || !sphb->ddw_enabled || windows ==
>>>> SPAPR_PCI_DMA_MAX_WINDOWS) {
>>>> + goto hw_error_exit;
>>>> + }
>>>> +
>>>> + tcet = spapr_tce_find_by_liobn(liobn);
>>>> + if (!tcet) {
>>>> + goto hw_error_exit;
>>>> + }
>>>> +
>>>> + win_addr = (windows == 0) ? sphb->dma_win_addr : sphb->dma64_win_addr;
>>>
>>> If the guest delets the default 32-bit window, then requests a really
>>> big 64-bit DMA window, will that work ok with the big window at 0
>>> instead of the usual 64-bit window address?
>>
>>
>> There is no valid guest to try that as they keep 32bit window.
>
> Right, but we should aim to work in general, not just with known
> guests.
Well, I did the experiment described below.
The term "in general" is vague though - if pHyp did something not exactly
as PAPR said (and therefore guests expected that), what behavior should I
pick for QEMU? For example, the guest did not expect a new window to start
from zero so there should have been reason for that, something like pHyp
only can allocate a single window and only at 1<<59 offset or nobody
actually tested it (always a possibility).
>> There was a relatively short period of time in v3.0-ish era (sles11 did
>> have it and sles11sp3 did not if I remember correctly) when the guest would
>> remove all windows and create one huge window but for some reason it
>> expected the window to start non from zero (perhaps pHyp implementation
>> detail) so it would fail. I did an experiment and removed that particular
>> check and it worked just fine.
>
> You mean removed the check for non-zero address from the guest?
Yes, that one.
>> Today guests always keep a 32bit window as the platform cannot tell if all
>> the drivers on a specific PHB will request 64bit DMA.
>>
>>
>>
>>
>>>> + spapr_tce_table_enable(tcet, page_shift, win_addr,
>>>> + 1ULL << (window_shift - page_shift));
>>>> + if (!tcet->nb_table) {
>>>> + goto hw_error_exit;
>>>> + }
>>>> +
>>>> + trace_spapr_iommu_ddw_create(buid, addr, 1ULL << page_shift,
>>>> + 1ULL << window_shift, tcet->bus_offset,
>>>> liobn);
>>>> +
>>>> + rtas_st(rets, 0, RTAS_OUT_SUCCESS);
>>>> + rtas_st(rets, 1, liobn);
>>>> + rtas_st(rets, 2, tcet->bus_offset >> 32);
>>>> + rtas_st(rets, 3, tcet->bus_offset & ((uint32_t) -1));
>>>> +
>>>> + return;
>>>> +
>>>> +hw_error_exit:
>>>> + rtas_st(rets, 0, RTAS_OUT_HW_ERROR);
>>>> + return;
>>>> +
>>>> +param_error_exit:
>>>> + rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
>>>> +}
>>>> +
>>>> +static void rtas_ibm_remove_pe_dma_window(PowerPCCPU *cpu,
>>>> + sPAPRMachineState *spapr,
>>>> + uint32_t token, uint32_t nargs,
>>>> + target_ulong args,
>>>> + uint32_t nret, target_ulong
>>>> rets)
>>>> +{
>>>> + sPAPRPHBState *sphb;
>>>> + sPAPRTCETable *tcet;
>>>> + uint32_t liobn;
>>>> +
>>>> + if ((nargs != 1) || (nret != 1)) {
>>>> + goto param_error_exit;
>>>> + }
>>>> +
>>>> + liobn = rtas_ld(args, 0);
>>>> + tcet = spapr_tce_find_by_liobn(liobn);
>>>> + if (!tcet) {
>>>> + goto param_error_exit;
>>>> + }
>>>> +
>>>> + sphb = SPAPR_PCI_HOST_BRIDGE(OBJECT(tcet)->parent);
>>>> + if (!sphb || !sphb->ddw_enabled || !tcet->nb_table) {
>>>> + goto param_error_exit;
>>>> + }
>>>> +
>>>> + spapr_tce_table_disable(tcet);
>>>> + trace_spapr_iommu_ddw_remove(liobn);
>>>> +
>>>> + rtas_st(rets, 0, RTAS_OUT_SUCCESS);
>>>> + return;
>>>> +
>>>> +param_error_exit:
>>>> + rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
>>>> +}
>>>> +
>>>> +static void rtas_ibm_reset_pe_dma_window(PowerPCCPU *cpu,
>>>> + sPAPRMachineState *spapr,
>>>> + uint32_t token, uint32_t nargs,
>>>> + target_ulong args,
>>>> + uint32_t nret, target_ulong rets)
>>>> +{
>>>> + sPAPRPHBState *sphb;
>>>> + uint64_t buid;
>>>> + uint32_t addr;
>>>> +
>>>> + if ((nargs != 3) || (nret != 1)) {
>>>> + goto param_error_exit;
>>>> + }
>>>> +
>>>> + buid = ((uint64_t)rtas_ld(args, 1) << 32) | rtas_ld(args, 2);
>>>> + addr = rtas_ld(args, 0);
>>>> + sphb = spapr_pci_find_phb(spapr, buid);
>>>> + if (!sphb || !sphb->ddw_enabled) {
>>>> + goto param_error_exit;
>>>> + }
>>>> +
>>>> + spapr_phb_dma_reset(sphb);
>>>> + trace_spapr_iommu_ddw_reset(buid, addr);
>>>> +
>>>> + rtas_st(rets, 0, RTAS_OUT_SUCCESS);
>>>> +
>>>> + return;
>>>> +
>>>> +param_error_exit:
>>>> + rtas_st(rets, 0, RTAS_OUT_PARAM_ERROR);
>>>> +}
>>>> +
>>>> +static void spapr_rtas_ddw_init(void)
>>>> +{
>>>> + spapr_rtas_register(RTAS_IBM_QUERY_PE_DMA_WINDOW,
>>>> + "ibm,query-pe-dma-window",
>>>> + rtas_ibm_query_pe_dma_window);
>>>> + spapr_rtas_register(RTAS_IBM_CREATE_PE_DMA_WINDOW,
>>>> + "ibm,create-pe-dma-window",
>>>> + rtas_ibm_create_pe_dma_window);
>>>> + spapr_rtas_register(RTAS_IBM_REMOVE_PE_DMA_WINDOW,
>>>> + "ibm,remove-pe-dma-window",
>>>> + rtas_ibm_remove_pe_dma_window);
>>>> + spapr_rtas_register(RTAS_IBM_RESET_PE_DMA_WINDOW,
>>>> + "ibm,reset-pe-dma-window",
>>>> + rtas_ibm_reset_pe_dma_window);
>>>> +}
>>>> +
>>>> +type_init(spapr_rtas_ddw_init)
>>>> diff --git a/include/hw/pci-host/spapr.h b/include/hw/pci-host/spapr.h
>>>> index 7848366..92aa610 100644
>>>> --- a/include/hw/pci-host/spapr.h
>>>> +++ b/include/hw/pci-host/spapr.h
>>>> @@ -32,6 +32,8 @@
>>>> #define SPAPR_PCI_HOST_BRIDGE(obj) \
>>>> OBJECT_CHECK(sPAPRPHBState, (obj), TYPE_SPAPR_PCI_HOST_BRIDGE)
>>>>
>>>> +#define SPAPR_PCI_DMA_MAX_WINDOWS 2
>>>> +
>>>> typedef struct sPAPRPHBState sPAPRPHBState;
>>>>
>>>> typedef struct spapr_pci_msi {
>>>> @@ -56,7 +58,7 @@ struct sPAPRPHBState {
>>>> hwaddr mem_win_addr, mem_win_size, io_win_addr, io_win_size;
>>>> MemoryRegion memwindow, iowindow, msiwindow;
>>>>
>>>> - uint32_t dma_liobn;
>>>> + uint32_t dma_liobn[SPAPR_PCI_DMA_MAX_WINDOWS];
>>>> hwaddr dma_win_addr, dma_win_size;
>>>> AddressSpace iommu_as;
>>>> MemoryRegion iommu_root;
>>>> @@ -71,6 +73,10 @@ struct sPAPRPHBState {
>>>> spapr_pci_msi_mig *msi_devs;
>>>>
>>>> QLIST_ENTRY(sPAPRPHBState) list;
>>>> +
>>>> + bool ddw_enabled;
>>>> + uint64_t page_size_mask;
>>>> + uint64_t dma64_win_addr;
>>>> };
>>>>
>>>> #define SPAPR_PCI_MAX_INDEX 255
>>>> diff --git a/include/hw/ppc/spapr.h b/include/hw/ppc/spapr.h
>>>> index e1f8274..36d1748 100644
>>>> --- a/include/hw/ppc/spapr.h
>>>> +++ b/include/hw/ppc/spapr.h
>>>> @@ -416,6 +416,16 @@ int spapr_allocate_irq_block(int num, bool lsi, bool
>>>> msi);
>>>> #define RTAS_OUT_NOT_AUTHORIZED -9002
>>>> #define RTAS_OUT_SYSPARM_PARAM_ERROR -9999
>>>>
>>>> +/* DDW pagesize mask values from ibm,query-pe-dma-window */
>>>> +#define RTAS_DDW_PGSIZE_4K 0x01
>>>> +#define RTAS_DDW_PGSIZE_64K 0x02
>>>> +#define RTAS_DDW_PGSIZE_16M 0x04
>>>> +#define RTAS_DDW_PGSIZE_32M 0x08
>>>> +#define RTAS_DDW_PGSIZE_64M 0x10
>>>> +#define RTAS_DDW_PGSIZE_128M 0x20
>>>> +#define RTAS_DDW_PGSIZE_256M 0x40
>>>> +#define RTAS_DDW_PGSIZE_16G 0x80
>>>> +
>>>> /* RTAS tokens */
>>>> #define RTAS_TOKEN_BASE 0x2000
>>>>
>>>> @@ -457,8 +467,12 @@ int spapr_allocate_irq_block(int num, bool lsi, bool
>>>> msi);
>>>> #define RTAS_IBM_SET_SLOT_RESET (RTAS_TOKEN_BASE + 0x23)
>>>> #define RTAS_IBM_CONFIGURE_PE (RTAS_TOKEN_BASE + 0x24)
>>>> #define RTAS_IBM_SLOT_ERROR_DETAIL (RTAS_TOKEN_BASE + 0x25)
>>>> +#define RTAS_IBM_QUERY_PE_DMA_WINDOW (RTAS_TOKEN_BASE + 0x26)
>>>> +#define RTAS_IBM_CREATE_PE_DMA_WINDOW (RTAS_TOKEN_BASE + 0x27)
>>>> +#define RTAS_IBM_REMOVE_PE_DMA_WINDOW (RTAS_TOKEN_BASE + 0x28)
>>>> +#define RTAS_IBM_RESET_PE_DMA_WINDOW (RTAS_TOKEN_BASE + 0x29)
>>>>
>>>> -#define RTAS_TOKEN_MAX (RTAS_TOKEN_BASE + 0x26)
>>>> +#define RTAS_TOKEN_MAX (RTAS_TOKEN_BASE + 0x2A)
>>>>
>>>> /* RTAS ibm,get-system-parameter token values */
>>>> #define RTAS_SYSPARM_SPLPAR_CHARACTERISTICS 20
>>>> diff --git a/trace-events b/trace-events
>>>> index 7e94d92..5b52634 100644
>>>> --- a/trace-events
>>>> +++ b/trace-events
>>>> @@ -1435,6 +1435,10 @@ spapr_iommu_xlate(uint64_t liobn, uint64_t ioba,
>>>> uint64_t tce, unsigned perm, un
>>>> spapr_iommu_new_table(uint64_t liobn, void *table, int fd)
>>>> "liobn=%"PRIx64" table=%p fd=%d"
>>>> spapr_iommu_pre_save(uint64_t liobn, uint32_t nb, uint64_t offs, uint32_t
>>>> ps) "liobn=%"PRIx64" %"PRIx32" bus_offset=%"PRIx64" ps=%"PRIu32
>>>> spapr_iommu_post_load(uint64_t liobn, uint32_t pre_nb, uint32_t post_nb,
>>>> uint64_t offs, uint32_t ps) "liobn=%"PRIx64" %"PRIx32" => %"PRIx32"
>>>> bus_offset=%"PRIx64" ps=%"PRIu32
>>>> +spapr_iommu_ddw_query(uint64_t buid, uint32_t cfgaddr, unsigned wa,
>>>> uint64_t win_size, uint32_t pgmask) "buid=%"PRIx64" addr=%"PRIx32", %u
>>>> windows available, max window size=%"PRIx64", mask=%"PRIx32
>>>> +spapr_iommu_ddw_create(uint64_t buid, uint32_t cfgaddr, uint64_t pg_size,
>>>> uint64_t req_size, uint64_t start, uint32_t liobn) "buid=%"PRIx64"
>>>> addr=%"PRIx32", page size=0x%"PRIx64", requested=0x%"PRIx64", start
>>>> addr=%"PRIx64", liobn=%"PRIx32
>>>> +spapr_iommu_ddw_remove(uint32_t liobn) "liobn=%"PRIx32
>>>> +spapr_iommu_ddw_reset(uint64_t buid, uint32_t cfgaddr) "buid=%"PRIx64"
>>>> addr=%"PRIx32
>>>>
>>>> # hw/ppc/ppc.c
>>>> ppc_tb_adjust(uint64_t offs1, uint64_t offs2, int64_t diff, int64_t
>>>> seconds) "adjusted from 0x%"PRIx64" to 0x%"PRIx64", diff %"PRId64"
>>>> (%"PRId64"s)"
>>>
>>
>>
>
>
>
>
--
Alexey
signature.asc
Description: OpenPGP digital signature
- [Qemu-devel] [PATCH qemu v18 3/5] vfio: Add host side DMA window capabilities, (continued)