Re: [PATCH] powerpc/powernv: remove dead npu-dma code

From: Alexey Kardashevskiy
Date: Thu Nov 01 2018 - 00:41:17 EST




On 31/10/2018 00:31, Christoph Hellwig wrote:
> This code has never been unused in the kernel since it was merged, and
> there has been no attempt that I could find to even submit users for
> it. Besides the general policy of not keep 1000+ lines of dead
> code, it helps cleaning up the DMA code and making powerpc user common
> infrastructure.
>
> This effectively reverts commit 5d2aa710 ("powerpc/powernv: Add support
> for Nvlink NPUs").
>
> Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> ---
>
> [this is a follow up to the brief maintainer summit discussion]
>
> arch/powerpc/include/asm/pci.h | 3 -
> arch/powerpc/include/asm/powernv.h | 23 -
> arch/powerpc/platforms/powernv/Makefile | 2 +-
> arch/powerpc/platforms/powernv/npu-dma.c | 989 ----------------------
> arch/powerpc/platforms/powernv/pci-ioda.c | 243 ------
> arch/powerpc/platforms/powernv/pci.h | 11 -
> 6 files changed, 1 insertion(+), 1270 deletions(-)
> delete mode 100644 arch/powerpc/platforms/powernv/npu-dma.c
>
> diff --git a/arch/powerpc/include/asm/pci.h b/arch/powerpc/include/asm/pci.h
> index 2af9ded80540..a01d2e3d6ff9 100644
> --- a/arch/powerpc/include/asm/pci.h
> +++ b/arch/powerpc/include/asm/pci.h
> @@ -127,7 +127,4 @@ extern void pcibios_scan_phb(struct pci_controller *hose);
>
> #endif /* __KERNEL__ */
>
> -extern struct pci_dev *pnv_pci_get_gpu_dev(struct pci_dev *npdev);
> -extern struct pci_dev *pnv_pci_get_npu_dev(struct pci_dev *gpdev, int index);
> -
> #endif /* __ASM_POWERPC_PCI_H */
> diff --git a/arch/powerpc/include/asm/powernv.h b/arch/powerpc/include/asm/powernv.h
> index 2f3ff7a27881..4848a6b3c6b2 100644
> --- a/arch/powerpc/include/asm/powernv.h
> +++ b/arch/powerpc/include/asm/powernv.h
> @@ -11,33 +11,10 @@
> #define _ASM_POWERNV_H
>
> #ifdef CONFIG_PPC_POWERNV
> -#define NPU2_WRITE 1
> extern void powernv_set_nmmu_ptcr(unsigned long ptcr);
> -extern struct npu_context *pnv_npu2_init_context(struct pci_dev *gpdev,
> - unsigned long flags,
> - void (*cb)(struct npu_context *, void *),
> - void *priv);
> -extern void pnv_npu2_destroy_context(struct npu_context *context,
> - struct pci_dev *gpdev);
> -extern int pnv_npu2_handle_fault(struct npu_context *context, uintptr_t *ea,
> - unsigned long *flags, unsigned long *status,
> - int count);
> -
> void pnv_tm_init(void);
> #else
> static inline void powernv_set_nmmu_ptcr(unsigned long ptcr) { }
> -static inline struct npu_context *pnv_npu2_init_context(struct pci_dev *gpdev,
> - unsigned long flags,
> - struct npu_context *(*cb)(struct npu_context *, void *),
> - void *priv) { return ERR_PTR(-ENODEV); }
> -static inline void pnv_npu2_destroy_context(struct npu_context *context,
> - struct pci_dev *gpdev) { }
> -
> -static inline int pnv_npu2_handle_fault(struct npu_context *context,
> - uintptr_t *ea, unsigned long *flags,
> - unsigned long *status, int count) {
> - return -ENODEV;
> -}
>
> static inline void pnv_tm_init(void) { }
> static inline void pnv_power9_force_smt4(void) { }
> diff --git a/arch/powerpc/platforms/powernv/Makefile b/arch/powerpc/platforms/powernv/Makefile
> index b540ce8eec55..2b13e9dd137c 100644
> --- a/arch/powerpc/platforms/powernv/Makefile
> +++ b/arch/powerpc/platforms/powernv/Makefile
> @@ -6,7 +6,7 @@ obj-y += opal-msglog.o opal-hmi.o opal-power.o opal-irqchip.o
> obj-y += opal-kmsg.o opal-powercap.o opal-psr.o opal-sensor-groups.o
>
> obj-$(CONFIG_SMP) += smp.o subcore.o subcore-asm.o
> -obj-$(CONFIG_PCI) += pci.o pci-ioda.o npu-dma.o pci-ioda-tce.o
> +obj-$(CONFIG_PCI) += pci.o pci-ioda.o pci-ioda-tce.o
> obj-$(CONFIG_CXL_BASE) += pci-cxl.o
> obj-$(CONFIG_EEH) += eeh-powernv.o
> obj-$(CONFIG_PPC_SCOM) += opal-xscom.o
> diff --git a/arch/powerpc/platforms/powernv/npu-dma.c b/arch/powerpc/platforms/powernv/npu-dma.c
> deleted file mode 100644
> index 6f60e0931922..000000000000
> --- a/arch/powerpc/platforms/powernv/npu-dma.c
> +++ /dev/null
> @@ -1,989 +0,0 @@
> -/*
> - * This file implements the DMA operations for NVLink devices. The NPU
> - * devices all point to the same iommu table as the parent PCI device.
> - *
> - * Copyright Alistair Popple, IBM Corporation 2015.
> - *
> - * This program is free software; you can redistribute it and/or
> - * modify it under the terms of version 2 of the GNU General Public
> - * License as published by the Free Software Foundation.
> - */
> -
> -#include <linux/slab.h>
> -#include <linux/mmu_notifier.h>
> -#include <linux/mmu_context.h>
> -#include <linux/of.h>
> -#include <linux/export.h>
> -#include <linux/pci.h>
> -#include <linux/memblock.h>
> -#include <linux/iommu.h>
> -#include <linux/sizes.h>
> -
> -#include <asm/debugfs.h>
> -#include <asm/tlb.h>
> -#include <asm/powernv.h>
> -#include <asm/reg.h>
> -#include <asm/opal.h>
> -#include <asm/io.h>
> -#include <asm/iommu.h>
> -#include <asm/pnv-pci.h>
> -#include <asm/msi_bitmap.h>
> -#include <asm/opal.h>
> -
> -#include "powernv.h"
> -#include "pci.h"
> -
> -#define npu_to_phb(x) container_of(x, struct pnv_phb, npu)
> -
> -/*
> - * spinlock to protect initialisation of an npu_context for a particular
> - * mm_struct.
> - */
> -static DEFINE_SPINLOCK(npu_context_lock);
> -
> -/*
> - * Other types of TCE cache invalidation are not functional in the
> - * hardware.
> - */
> -static struct pci_dev *get_pci_dev(struct device_node *dn)
> -{
> - struct pci_dn *pdn = PCI_DN(dn);
> -
> - return pci_get_domain_bus_and_slot(pci_domain_nr(pdn->phb->bus),
> - pdn->busno, pdn->devfn);
> -}
> -
> -/* Given a NPU device get the associated PCI device. */
> -struct pci_dev *pnv_pci_get_gpu_dev(struct pci_dev *npdev)
> -{
> - struct device_node *dn;
> - struct pci_dev *gpdev;
> -
> - if (WARN_ON(!npdev))
> - return NULL;
> -
> - if (WARN_ON(!npdev->dev.of_node))
> - return NULL;
> -
> - /* Get assoicated PCI device */
> - dn = of_parse_phandle(npdev->dev.of_node, "ibm,gpu", 0);
> - if (!dn)
> - return NULL;
> -
> - gpdev = get_pci_dev(dn);
> - of_node_put(dn);
> -
> - return gpdev;
> -}
> -EXPORT_SYMBOL(pnv_pci_get_gpu_dev);
> -
> -/* Given the real PCI device get a linked NPU device. */
> -struct pci_dev *pnv_pci_get_npu_dev(struct pci_dev *gpdev, int index)
> -{
> - struct device_node *dn;
> - struct pci_dev *npdev;
> -
> - if (WARN_ON(!gpdev))
> - return NULL;
> -
> - /* Not all PCI devices have device-tree nodes */
> - if (!gpdev->dev.of_node)
> - return NULL;
> -
> - /* Get assoicated PCI device */
> - dn = of_parse_phandle(gpdev->dev.of_node, "ibm,npu", index);
> - if (!dn)
> - return NULL;
> -
> - npdev = get_pci_dev(dn);
> - of_node_put(dn);
> -
> - return npdev;
> -}
> -EXPORT_SYMBOL(pnv_pci_get_npu_dev);
> -
> -#define NPU_DMA_OP_UNSUPPORTED() \
> - dev_err_once(dev, "%s operation unsupported for NVLink devices\n", \
> - __func__)
> -
> -static void *dma_npu_alloc(struct device *dev, size_t size,
> - dma_addr_t *dma_handle, gfp_t flag,
> - unsigned long attrs)
> -{
> - NPU_DMA_OP_UNSUPPORTED();
> - return NULL;
> -}
> -
> -static void dma_npu_free(struct device *dev, size_t size,
> - void *vaddr, dma_addr_t dma_handle,
> - unsigned long attrs)
> -{
> - NPU_DMA_OP_UNSUPPORTED();
> -}
> -
> -static dma_addr_t dma_npu_map_page(struct device *dev, struct page *page,
> - unsigned long offset, size_t size,
> - enum dma_data_direction direction,
> - unsigned long attrs)
> -{
> - NPU_DMA_OP_UNSUPPORTED();
> - return 0;
> -}
> -
> -static int dma_npu_map_sg(struct device *dev, struct scatterlist *sglist,
> - int nelems, enum dma_data_direction direction,
> - unsigned long attrs)
> -{
> - NPU_DMA_OP_UNSUPPORTED();
> - return 0;
> -}
> -
> -static int dma_npu_dma_supported(struct device *dev, u64 mask)
> -{
> - NPU_DMA_OP_UNSUPPORTED();
> - return 0;
> -}
> -
> -static u64 dma_npu_get_required_mask(struct device *dev)
> -{
> - NPU_DMA_OP_UNSUPPORTED();
> - return 0;
> -}
> -
> -static const struct dma_map_ops dma_npu_ops = {
> - .map_page = dma_npu_map_page,
> - .map_sg = dma_npu_map_sg,
> - .alloc = dma_npu_alloc,
> - .free = dma_npu_free,
> - .dma_supported = dma_npu_dma_supported,
> - .get_required_mask = dma_npu_get_required_mask,
> -};
> -
> -/*
> - * Returns the PE assoicated with the PCI device of the given
> - * NPU. Returns the linked pci device if pci_dev != NULL.
> - */
> -static struct pnv_ioda_pe *get_gpu_pci_dev_and_pe(struct pnv_ioda_pe *npe,
> - struct pci_dev **gpdev)
> -{
> - struct pnv_phb *phb;
> - struct pci_controller *hose;
> - struct pci_dev *pdev;
> - struct pnv_ioda_pe *pe;
> - struct pci_dn *pdn;
> -
> - pdev = pnv_pci_get_gpu_dev(npe->pdev);
> - if (!pdev)
> - return NULL;
> -
> - pdn = pci_get_pdn(pdev);
> - if (WARN_ON(!pdn || pdn->pe_number == IODA_INVALID_PE))
> - return NULL;
> -
> - hose = pci_bus_to_host(pdev->bus);
> - phb = hose->private_data;
> - pe = &phb->ioda.pe_array[pdn->pe_number];
> -
> - if (gpdev)
> - *gpdev = pdev;
> -
> - return pe;
> -}
> -
> -long pnv_npu_set_window(struct pnv_ioda_pe *npe, int num,
> - struct iommu_table *tbl)
> -{
> - struct pnv_phb *phb = npe->phb;
> - int64_t rc;
> - const unsigned long size = tbl->it_indirect_levels ?
> - tbl->it_level_size : tbl->it_size;
> - const __u64 start_addr = tbl->it_offset << tbl->it_page_shift;
> - const __u64 win_size = tbl->it_size << tbl->it_page_shift;
> -
> - pe_info(npe, "Setting up window %llx..%llx pg=%lx\n",
> - start_addr, start_addr + win_size - 1,
> - IOMMU_PAGE_SIZE(tbl));
> -
> - rc = opal_pci_map_pe_dma_window(phb->opal_id,
> - npe->pe_number,
> - npe->pe_number,
> - tbl->it_indirect_levels + 1,
> - __pa(tbl->it_base),
> - size << 3,
> - IOMMU_PAGE_SIZE(tbl));
> - if (rc) {
> - pe_err(npe, "Failed to configure TCE table, err %lld\n", rc);
> - return rc;
> - }
> - pnv_pci_ioda2_tce_invalidate_entire(phb, false);
> -
> - /* Add the table to the list so its TCE cache will get invalidated */
> - pnv_pci_link_table_and_group(phb->hose->node, num,
> - tbl, &npe->table_group);
> -
> - return 0;
> -}
> -
> -long pnv_npu_unset_window(struct pnv_ioda_pe *npe, int num)
> -{
> - struct pnv_phb *phb = npe->phb;
> - int64_t rc;
> -
> - pe_info(npe, "Removing DMA window\n");
> -
> - rc = opal_pci_map_pe_dma_window(phb->opal_id, npe->pe_number,
> - npe->pe_number,
> - 0/* levels */, 0/* table address */,
> - 0/* table size */, 0/* page size */);
> - if (rc) {
> - pe_err(npe, "Unmapping failed, ret = %lld\n", rc);
> - return rc;
> - }
> - pnv_pci_ioda2_tce_invalidate_entire(phb, false);
> -
> - pnv_pci_unlink_table_and_group(npe->table_group.tables[num],
> - &npe->table_group);
> -
> - return 0;
> -}
> -
> -/*
> - * Enables 32 bit DMA on NPU.
> - */
> -static void pnv_npu_dma_set_32(struct pnv_ioda_pe *npe)
> -{
> - struct pci_dev *gpdev;
> - struct pnv_ioda_pe *gpe;
> - int64_t rc;
> -
> - /*
> - * Find the assoicated PCI devices and get the dma window
> - * information from there.
> - */
> - if (!npe->pdev || !(npe->flags & PNV_IODA_PE_DEV))
> - return;
> -
> - gpe = get_gpu_pci_dev_and_pe(npe, &gpdev);
> - if (!gpe)
> - return;
> -
> - rc = pnv_npu_set_window(npe, 0, gpe->table_group.tables[0]);
> -
> - /*
> - * We don't initialise npu_pe->tce32_table as we always use
> - * dma_npu_ops which are nops.
> - */
> - set_dma_ops(&npe->pdev->dev, &dma_npu_ops);
> -}
> -
> -/*
> - * Enables bypass mode on the NPU. The NPU only supports one
> - * window per link, so bypass needs to be explicitly enabled or
> - * disabled. Unlike for a PHB3 bypass and non-bypass modes can't be
> - * active at the same time.
> - */
> -static int pnv_npu_dma_set_bypass(struct pnv_ioda_pe *npe)
> -{
> - struct pnv_phb *phb = npe->phb;
> - int64_t rc = 0;
> - phys_addr_t top = memblock_end_of_DRAM();
> -
> - if (phb->type != PNV_PHB_NPU_NVLINK || !npe->pdev)
> - return -EINVAL;
> -
> - rc = pnv_npu_unset_window(npe, 0);
> - if (rc != OPAL_SUCCESS)
> - return rc;
> -
> - /* Enable the bypass window */
> -
> - top = roundup_pow_of_two(top);
> - dev_info(&npe->pdev->dev, "Enabling bypass for PE %x\n",
> - npe->pe_number);
> - rc = opal_pci_map_pe_dma_window_real(phb->opal_id,
> - npe->pe_number, npe->pe_number,
> - 0 /* bypass base */, top);
> -
> - if (rc == OPAL_SUCCESS)
> - pnv_pci_ioda2_tce_invalidate_entire(phb, false);
> -
> - return rc;
> -}
> -
> -void pnv_npu_try_dma_set_bypass(struct pci_dev *gpdev, bool bypass)
> -{
> - int i;
> - struct pnv_phb *phb;
> - struct pci_dn *pdn;
> - struct pnv_ioda_pe *npe;
> - struct pci_dev *npdev;
> -
> - for (i = 0; ; ++i) {
> - npdev = pnv_pci_get_npu_dev(gpdev, i);
> -
> - if (!npdev)
> - break;
> -
> - pdn = pci_get_pdn(npdev);
> - if (WARN_ON(!pdn || pdn->pe_number == IODA_INVALID_PE))
> - return;
> -
> - phb = pci_bus_to_host(npdev->bus)->private_data;
> -
> - /* We only do bypass if it's enabled on the linked device */
> - npe = &phb->ioda.pe_array[pdn->pe_number];
> -
> - if (bypass) {
> - dev_info(&npdev->dev,
> - "Using 64-bit DMA iommu bypass\n");
> - pnv_npu_dma_set_bypass(npe);
> - } else {
> - dev_info(&npdev->dev, "Using 32-bit DMA via iommu\n");
> - pnv_npu_dma_set_32(npe);
> - }
> - }
> -}
> -
> -/* Switch ownership from platform code to external user (e.g. VFIO) */
> -void pnv_npu_take_ownership(struct pnv_ioda_pe *npe)
> -{
> - struct pnv_phb *phb = npe->phb;
> - int64_t rc;
> -
> - /*
> - * Note: NPU has just a single TVE in the hardware which means that
> - * while used by the kernel, it can have either 32bit window or
> - * DMA bypass but never both. So we deconfigure 32bit window only
> - * if it was enabled at the moment of ownership change.
> - */
> - if (npe->table_group.tables[0]) {
> - pnv_npu_unset_window(npe, 0);
> - return;
> - }
> -
> - /* Disable bypass */
> - rc = opal_pci_map_pe_dma_window_real(phb->opal_id,
> - npe->pe_number, npe->pe_number,
> - 0 /* bypass base */, 0);
> - if (rc) {
> - pe_err(npe, "Failed to disable bypass, err %lld\n", rc);
> - return;
> - }
> - pnv_pci_ioda2_tce_invalidate_entire(npe->phb, false);
> -}
> -
> -struct pnv_ioda_pe *pnv_pci_npu_setup_iommu(struct pnv_ioda_pe *npe)
> -{
> - struct pnv_phb *phb = npe->phb;
> - struct pci_bus *pbus = phb->hose->bus;
> - struct pci_dev *npdev, *gpdev = NULL, *gptmp;
> - struct pnv_ioda_pe *gpe = get_gpu_pci_dev_and_pe(npe, &gpdev);
> -
> - if (!gpe || !gpdev)
> - return NULL;
> -
> - list_for_each_entry(npdev, &pbus->devices, bus_list) {
> - gptmp = pnv_pci_get_gpu_dev(npdev);
> -
> - if (gptmp != gpdev)
> - continue;
> -
> - pe_info(gpe, "Attached NPU %s\n", dev_name(&npdev->dev));
> - iommu_group_add_device(gpe->table_group.group, &npdev->dev);
> - }
> -
> - return gpe;
> -}
> -
> -/* Maximum number of nvlinks per npu */
> -#define NV_MAX_LINKS 6
> -
> -/* Maximum index of npu2 hosts in the system. Always < NV_MAX_NPUS */
> -static int max_npu2_index;
> -
> -struct npu_context {
> - struct mm_struct *mm;
> - struct pci_dev *npdev[NV_MAX_NPUS][NV_MAX_LINKS];
> - struct mmu_notifier mn;
> - struct kref kref;
> - bool nmmu_flush;
> -
> - /* Callback to stop translation requests on a given GPU */
> - void (*release_cb)(struct npu_context *context, void *priv);
> -
> - /*
> - * Private pointer passed to the above callback for usage by
> - * device drivers.
> - */
> - void *priv;
> -};
> -
> -struct mmio_atsd_reg {
> - struct npu *npu;
> - int reg;
> -};
> -
> -/*
> - * Find a free MMIO ATSD register and mark it in use. Return -ENOSPC
> - * if none are available.
> - */
> -static int get_mmio_atsd_reg(struct npu *npu)
> -{
> - int i;
> -
> - for (i = 0; i < npu->mmio_atsd_count; i++) {
> - if (!test_bit(i, &npu->mmio_atsd_usage))
> - if (!test_and_set_bit_lock(i, &npu->mmio_atsd_usage))
> - return i;
> - }
> -
> - return -ENOSPC;
> -}
> -
> -static void put_mmio_atsd_reg(struct npu *npu, int reg)
> -{
> - clear_bit_unlock(reg, &npu->mmio_atsd_usage);
> -}
> -
> -/* MMIO ATSD register offsets */
> -#define XTS_ATSD_LAUNCH 0
> -#define XTS_ATSD_AVA 1
> -#define XTS_ATSD_STAT 2
> -
> -static unsigned long get_atsd_launch_val(unsigned long pid, unsigned long psize)
> -{
> - unsigned long launch = 0;
> -
> - if (psize == MMU_PAGE_COUNT) {
> - /* IS set to invalidate entire matching PID */
> - launch |= PPC_BIT(12);
> - } else {
> - /* AP set to invalidate region of psize */
> - launch |= (u64)mmu_get_ap(psize) << PPC_BITLSHIFT(17);
> - }
> -
> - /* PRS set to process-scoped */
> - launch |= PPC_BIT(13);
> -
> - /* PID */
> - launch |= pid << PPC_BITLSHIFT(38);
> -
> - /* Leave "No flush" (bit 39) 0 so every ATSD performs a flush */
> -
> - return launch;
> -}
> -
> -static void mmio_atsd_regs_write(struct mmio_atsd_reg
> - mmio_atsd_reg[NV_MAX_NPUS], unsigned long offset,
> - unsigned long val)
> -{
> - struct npu *npu;
> - int i, reg;
> -
> - for (i = 0; i <= max_npu2_index; i++) {
> - reg = mmio_atsd_reg[i].reg;
> - if (reg < 0)
> - continue;
> -
> - npu = mmio_atsd_reg[i].npu;
> - __raw_writeq_be(val, npu->mmio_atsd_regs[reg] + offset);
> - }
> -}
> -
> -static void mmio_invalidate_pid(struct mmio_atsd_reg mmio_atsd_reg[NV_MAX_NPUS],
> - unsigned long pid)
> -{
> - unsigned long launch = get_atsd_launch_val(pid, MMU_PAGE_COUNT);
> -
> - /* Invalidating the entire process doesn't use a va */
> - mmio_atsd_regs_write(mmio_atsd_reg, XTS_ATSD_LAUNCH, launch);
> -}
> -
> -static void mmio_invalidate_range(struct mmio_atsd_reg
> - mmio_atsd_reg[NV_MAX_NPUS], unsigned long pid,
> - unsigned long start, unsigned long psize)
> -{
> - unsigned long launch = get_atsd_launch_val(pid, psize);
> -
> - /* Write all VAs first */
> - mmio_atsd_regs_write(mmio_atsd_reg, XTS_ATSD_AVA, start);
> -
> - /* Issue one barrier for all address writes */
> - eieio();
> -
> - /* Launch */
> - mmio_atsd_regs_write(mmio_atsd_reg, XTS_ATSD_LAUNCH, launch);
> -}
> -
> -#define mn_to_npu_context(x) container_of(x, struct npu_context, mn)
> -
> -static void mmio_invalidate_wait(
> - struct mmio_atsd_reg mmio_atsd_reg[NV_MAX_NPUS])
> -{
> - struct npu *npu;
> - int i, reg;
> -
> - /* Wait for all invalidations to complete */
> - for (i = 0; i <= max_npu2_index; i++) {
> - if (mmio_atsd_reg[i].reg < 0)
> - continue;
> -
> - /* Wait for completion */
> - npu = mmio_atsd_reg[i].npu;
> - reg = mmio_atsd_reg[i].reg;
> - while (__raw_readq(npu->mmio_atsd_regs[reg] + XTS_ATSD_STAT))
> - cpu_relax();
> - }
> -}
> -
> -/*
> - * Acquires all the address translation shootdown (ATSD) registers required to
> - * launch an ATSD on all links this npu_context is active on.
> - */
> -static void acquire_atsd_reg(struct npu_context *npu_context,
> - struct mmio_atsd_reg mmio_atsd_reg[NV_MAX_NPUS])
> -{
> - int i, j;
> - struct npu *npu;
> - struct pci_dev *npdev;
> - struct pnv_phb *nphb;
> -
> - for (i = 0; i <= max_npu2_index; i++) {
> - mmio_atsd_reg[i].reg = -1;
> - for (j = 0; j < NV_MAX_LINKS; j++) {
> - /*
> - * There are no ordering requirements with respect to
> - * the setup of struct npu_context, but to ensure
> - * consistent behaviour we need to ensure npdev[][] is
> - * only read once.
> - */
> - npdev = READ_ONCE(npu_context->npdev[i][j]);
> - if (!npdev)
> - continue;
> -
> - nphb = pci_bus_to_host(npdev->bus)->private_data;
> - npu = &nphb->npu;
> - mmio_atsd_reg[i].npu = npu;
> - mmio_atsd_reg[i].reg = get_mmio_atsd_reg(npu);
> - while (mmio_atsd_reg[i].reg < 0) {
> - mmio_atsd_reg[i].reg = get_mmio_atsd_reg(npu);
> - cpu_relax();
> - }
> - break;
> - }
> - }
> -}
> -
> -/*
> - * Release previously acquired ATSD registers. To avoid deadlocks the registers
> - * must be released in the same order they were acquired above in
> - * acquire_atsd_reg.
> - */
> -static void release_atsd_reg(struct mmio_atsd_reg mmio_atsd_reg[NV_MAX_NPUS])
> -{
> - int i;
> -
> - for (i = 0; i <= max_npu2_index; i++) {
> - /*
> - * We can't rely on npu_context->npdev[][] being the same here
> - * as when acquire_atsd_reg() was called, hence we use the
> - * values stored in mmio_atsd_reg during the acquire phase
> - * rather than re-reading npdev[][].
> - */
> - if (mmio_atsd_reg[i].reg < 0)
> - continue;
> -
> - put_mmio_atsd_reg(mmio_atsd_reg[i].npu, mmio_atsd_reg[i].reg);
> - }
> -}
> -
> -/*
> - * Invalidate a virtual address range
> - */
> -static void mmio_invalidate(struct npu_context *npu_context,
> - unsigned long start, unsigned long size)
> -{
> - struct mmio_atsd_reg mmio_atsd_reg[NV_MAX_NPUS];
> - unsigned long pid = npu_context->mm->context.id;
> - unsigned long atsd_start = 0;
> - unsigned long end = start + size - 1;
> - int atsd_psize = MMU_PAGE_COUNT;
> -
> - /*
> - * Convert the input range into one of the supported sizes. If the range
> - * doesn't fit, use the next larger supported size. Invalidation latency
> - * is high, so over-invalidation is preferred to issuing multiple
> - * invalidates.
> - *
> - * A 4K page size isn't supported by NPU/GPU ATS, so that case is
> - * ignored.
> - */
> - if (size == SZ_64K) {
> - atsd_start = start;
> - atsd_psize = MMU_PAGE_64K;
> - } else if (ALIGN_DOWN(start, SZ_2M) == ALIGN_DOWN(end, SZ_2M)) {
> - atsd_start = ALIGN_DOWN(start, SZ_2M);
> - atsd_psize = MMU_PAGE_2M;
> - } else if (ALIGN_DOWN(start, SZ_1G) == ALIGN_DOWN(end, SZ_1G)) {
> - atsd_start = ALIGN_DOWN(start, SZ_1G);
> - atsd_psize = MMU_PAGE_1G;
> - }
> -
> - if (npu_context->nmmu_flush)
> - /*
> - * Unfortunately the nest mmu does not support flushing specific
> - * addresses so we have to flush the whole mm once before
> - * shooting down the GPU translation.
> - */
> - flush_all_mm(npu_context->mm);
> -
> - /*
> - * Loop over all the NPUs this process is active on and launch
> - * an invalidate.
> - */
> - acquire_atsd_reg(npu_context, mmio_atsd_reg);
> -
> - if (atsd_psize == MMU_PAGE_COUNT)
> - mmio_invalidate_pid(mmio_atsd_reg, pid);
> - else
> - mmio_invalidate_range(mmio_atsd_reg, pid, atsd_start,
> - atsd_psize);
> -
> - mmio_invalidate_wait(mmio_atsd_reg);
> -
> - /*
> - * The GPU requires two flush ATSDs to ensure all entries have been
> - * flushed. We use PID 0 as it will never be used for a process on the
> - * GPU.
> - */
> - mmio_invalidate_pid(mmio_atsd_reg, 0);
> - mmio_invalidate_wait(mmio_atsd_reg);
> - mmio_invalidate_pid(mmio_atsd_reg, 0);
> - mmio_invalidate_wait(mmio_atsd_reg);
> -
> - release_atsd_reg(mmio_atsd_reg);
> -}
> -
> -static void pnv_npu2_mn_release(struct mmu_notifier *mn,
> - struct mm_struct *mm)
> -{
> - struct npu_context *npu_context = mn_to_npu_context(mn);
> -
> - /* Call into device driver to stop requests to the NMMU */
> - if (npu_context->release_cb)
> - npu_context->release_cb(npu_context, npu_context->priv);
> -
> - /*
> - * There should be no more translation requests for this PID, but we
> - * need to ensure any entries for it are removed from the TLB.
> - */
> - mmio_invalidate(npu_context, 0, ~0UL);
> -}
> -
> -static void pnv_npu2_mn_change_pte(struct mmu_notifier *mn,
> - struct mm_struct *mm,
> - unsigned long address,
> - pte_t pte)
> -{
> - struct npu_context *npu_context = mn_to_npu_context(mn);
> - mmio_invalidate(npu_context, address, PAGE_SIZE);
> -}
> -
> -static void pnv_npu2_mn_invalidate_range(struct mmu_notifier *mn,
> - struct mm_struct *mm,
> - unsigned long start, unsigned long end)
> -{
> - struct npu_context *npu_context = mn_to_npu_context(mn);
> - mmio_invalidate(npu_context, start, end - start);
> -}
> -
> -static const struct mmu_notifier_ops nv_nmmu_notifier_ops = {
> - .release = pnv_npu2_mn_release,
> - .change_pte = pnv_npu2_mn_change_pte,
> - .invalidate_range = pnv_npu2_mn_invalidate_range,
> -};
> -
> -/*
> - * Call into OPAL to setup the nmmu context for the current task in
> - * the NPU. This must be called to setup the context tables before the
> - * GPU issues ATRs. pdev should be a pointed to PCIe GPU device.
> - *
> - * A release callback should be registered to allow a device driver to
> - * be notified that it should not launch any new translation requests
> - * as the final TLB invalidate is about to occur.
> - *
> - * Returns an error if there no contexts are currently available or a
> - * npu_context which should be passed to pnv_npu2_handle_fault().
> - *
> - * mmap_sem must be held in write mode and must not be called from interrupt
> - * context.
> - */
> -struct npu_context *pnv_npu2_init_context(struct pci_dev *gpdev,
> - unsigned long flags,
> - void (*cb)(struct npu_context *, void *),
> - void *priv)

> -{
> - int rc;
> - u32 nvlink_index;
> - struct device_node *nvlink_dn;
> - struct mm_struct *mm = current->mm;
> - struct pnv_phb *nphb;
> - struct npu *npu;
> - struct npu_context *npu_context;
> -
> - /*
> - * At present we don't support GPUs connected to multiple NPUs and I'm
> - * not sure the hardware does either.
> - */
> - struct pci_dev *npdev = pnv_pci_get_npu_dev(gpdev, 0);
> -
> - if (!firmware_has_feature(FW_FEATURE_OPAL))
> - return ERR_PTR(-ENODEV);
> -
> - if (!npdev)
> - /* No nvlink associated with this GPU device */
> - return ERR_PTR(-ENODEV);
> -
> - nvlink_dn = of_parse_phandle(npdev->dev.of_node, "ibm,nvlink", 0);
> - if (WARN_ON(of_property_read_u32(nvlink_dn, "ibm,npu-link-index",
> - &nvlink_index)))
> - return ERR_PTR(-ENODEV);
> -
> - if (!mm || mm->context.id == 0) {
> - /*
> - * Kernel thread contexts are not supported and context id 0 is
> - * reserved on the GPU.
> - */
> - return ERR_PTR(-EINVAL);
> - }
> -
> - nphb = pci_bus_to_host(npdev->bus)->private_data;
> - npu = &nphb->npu;
> -
> - /*
> - * Setup the NPU context table for a particular GPU. These need to be
> - * per-GPU as we need the tables to filter ATSDs when there are no
> - * active contexts on a particular GPU. It is safe for these to be
> - * called concurrently with destroy as the OPAL call takes appropriate
> - * locks and refcounts on init/destroy.
> - */
> - rc = opal_npu_init_context(nphb->opal_id, mm->context.id, flags,
> - PCI_DEVID(gpdev->bus->number, gpdev->devfn));
> - if (rc < 0)
> - return ERR_PTR(-ENOSPC);
> -
> - /*
> - * We store the npu pci device so we can more easily get at the
> - * associated npus.
> - */
> - spin_lock(&npu_context_lock);
> - npu_context = mm->context.npu_context;
> - if (npu_context) {
> - if (npu_context->release_cb != cb ||
> - npu_context->priv != priv) {
> - spin_unlock(&npu_context_lock);
> - opal_npu_destroy_context(nphb->opal_id, mm->context.id,
> - PCI_DEVID(gpdev->bus->number,
> - gpdev->devfn));
> - return ERR_PTR(-EINVAL);
> - }
> -
> - WARN_ON(!kref_get_unless_zero(&npu_context->kref));
> - }
> - spin_unlock(&npu_context_lock);
> -
> - if (!npu_context) {
> - /*
> - * We can set up these fields without holding the
> - * npu_context_lock as the npu_context hasn't been returned to
> - * the caller meaning it can't be destroyed. Parallel allocation
> - * is protected against by mmap_sem.
> - */
> - rc = -ENOMEM;
> - npu_context = kzalloc(sizeof(struct npu_context), GFP_KERNEL);
> - if (npu_context) {
> - kref_init(&npu_context->kref);
> - npu_context->mm = mm;
> - npu_context->mn.ops = &nv_nmmu_notifier_ops;
> - rc = __mmu_notifier_register(&npu_context->mn, mm);
> - }
> -
> - if (rc) {
> - kfree(npu_context);
> - opal_npu_destroy_context(nphb->opal_id, mm->context.id,
> - PCI_DEVID(gpdev->bus->number,
> - gpdev->devfn));
> - return ERR_PTR(rc);
> - }
> -
> - mm->context.npu_context = npu_context;
> - }
> -
> - npu_context->release_cb = cb;
> - npu_context->priv = priv;
> -
> - /*
> - * npdev is a pci_dev pointer setup by the PCI code. We assign it to
> - * npdev[][] to indicate to the mmu notifiers that an invalidation
> - * should also be sent over this nvlink. The notifiers don't use any
> - * other fields in npu_context, so we just need to ensure that when they
> - * deference npu_context->npdev[][] it is either a valid pointer or
> - * NULL.
> - */
> - WRITE_ONCE(npu_context->npdev[npu->index][nvlink_index], npdev);
> -
> - if (!nphb->npu.nmmu_flush) {
> - /*
> - * If we're not explicitly flushing ourselves we need to mark
> - * the thread for global flushes
> - */
> - npu_context->nmmu_flush = false;
> - mm_context_add_copro(mm);
> - } else
> - npu_context->nmmu_flush = true;
> -
> - return npu_context;
> -}
> -EXPORT_SYMBOL(pnv_npu2_init_context);
> -
> -static void pnv_npu2_release_context(struct kref *kref)
> -{
> - struct npu_context *npu_context =
> - container_of(kref, struct npu_context, kref);
> -
> - if (!npu_context->nmmu_flush)
> - mm_context_remove_copro(npu_context->mm);
> -
> - npu_context->mm->context.npu_context = NULL;
> -}
> -
> -/*
> - * Destroy a context on the given GPU. May free the npu_context if it is no
> - * longer active on any GPUs. Must not be called from interrupt context.
> - */
> -void pnv_npu2_destroy_context(struct npu_context *npu_context,
> - struct pci_dev *gpdev)
> -{
> - int removed;
> - struct pnv_phb *nphb;
> - struct npu *npu;
> - struct pci_dev *npdev = pnv_pci_get_npu_dev(gpdev, 0);
> - struct device_node *nvlink_dn;
> - u32 nvlink_index;
> -
> - if (WARN_ON(!npdev))
> - return;
> -
> - if (!firmware_has_feature(FW_FEATURE_OPAL))
> - return;
> -
> - nphb = pci_bus_to_host(npdev->bus)->private_data;
> - npu = &nphb->npu;
> - nvlink_dn = of_parse_phandle(npdev->dev.of_node, "ibm,nvlink", 0);
> - if (WARN_ON(of_property_read_u32(nvlink_dn, "ibm,npu-link-index",
> - &nvlink_index)))
> - return;
> - WRITE_ONCE(npu_context->npdev[npu->index][nvlink_index], NULL);
> - opal_npu_destroy_context(nphb->opal_id, npu_context->mm->context.id,
> - PCI_DEVID(gpdev->bus->number, gpdev->devfn));
> - spin_lock(&npu_context_lock);
> - removed = kref_put(&npu_context->kref, pnv_npu2_release_context);
> - spin_unlock(&npu_context_lock);
> -
> - /*
> - * We need to do this outside of pnv_npu2_release_context so that it is
> - * outside the spinlock as mmu_notifier_destroy uses SRCU.
> - */
> - if (removed) {
> - mmu_notifier_unregister(&npu_context->mn,
> - npu_context->mm);
> -
> - kfree(npu_context);
> - }
> -
> -}
> -EXPORT_SYMBOL(pnv_npu2_destroy_context);
> -
> -/*
> - * Assumes mmap_sem is held for the contexts associated mm.
> - */
> -int pnv_npu2_handle_fault(struct npu_context *context, uintptr_t *ea,
> - unsigned long *flags, unsigned long *status, int count)
> -{
> - u64 rc = 0, result = 0;
> - int i, is_write;
> - struct page *page[1];
> -
> - /* mmap_sem should be held so the struct_mm must be present */
> - struct mm_struct *mm = context->mm;
> -
> - if (!firmware_has_feature(FW_FEATURE_OPAL))
> - return -ENODEV;
> -
> - WARN_ON(!rwsem_is_locked(&mm->mmap_sem));
> -
> - for (i = 0; i < count; i++) {
> - is_write = flags[i] & NPU2_WRITE;
> - rc = get_user_pages_remote(NULL, mm, ea[i], 1,
> - is_write ? FOLL_WRITE : 0,
> - page, NULL, NULL);
> -
> - /*
> - * To support virtualised environments we will have to do an
> - * access to the page to ensure it gets faulted into the
> - * hypervisor. For the moment virtualisation is not supported in
> - * other areas so leave the access out.
> - */
> - if (rc != 1) {
> - status[i] = rc;
> - result = -EFAULT;
> - continue;
> - }
> -
> - status[i] = 0;
> - put_page(page[0]);
> - }
> -
> - return result;
> -}
> -EXPORT_SYMBOL(pnv_npu2_handle_fault);
> -
> -int pnv_npu2_init(struct pnv_phb *phb)
> -{
> - unsigned int i;
> - u64 mmio_atsd;
> - struct device_node *dn;
> - struct pci_dev *gpdev;
> - static int npu_index;
> - uint64_t rc = 0;
> -
> - phb->npu.nmmu_flush =
> - of_property_read_bool(phb->hose->dn, "ibm,nmmu-flush");
> - for_each_child_of_node(phb->hose->dn, dn) {
> - gpdev = pnv_pci_get_gpu_dev(get_pci_dev(dn));
> - if (gpdev) {
> - rc = opal_npu_map_lpar(phb->opal_id,
> - PCI_DEVID(gpdev->bus->number, gpdev->devfn),
> - 0, 0);

I heavily hacked on these (draft):
https://patchwork.ozlabs.org/project/kvm-ppc/list/?series=70791

and the result is going to be used for VFIO purposes (draft too):
https://patchwork.ozlabs.org/project/kvm-ppc/list/?series=70793

Other bits can go away from npu-dma but I still need some to make
KVM+VFIO work for GPUs, this is without any vendor driver on the host
side whatsoever.



> - if (rc)
> - dev_err(&gpdev->dev,
> - "Error %lld mapping device to LPAR\n",
> - rc);
> - }
> - }
> -
> - for (i = 0; !of_property_read_u64_index(phb->hose->dn, "ibm,mmio-atsd",
> - i, &mmio_atsd); i++)
> - phb->npu.mmio_atsd_regs[i] = ioremap(mmio_atsd, 32);
> -
> - pr_info("NPU%lld: Found %d MMIO ATSD registers", phb->opal_id, i);
> - phb->npu.mmio_atsd_count = i;
> - phb->npu.mmio_atsd_usage = 0;
> - npu_index++;
> - if (WARN_ON(npu_index >= NV_MAX_NPUS))
> - return -ENOSPC;
> - max_npu2_index = npu_index;
> - phb->npu.index = npu_index;
> -
> - return 0;
> -}
> diff --git a/arch/powerpc/platforms/powernv/pci-ioda.c b/arch/powerpc/platforms/powernv/pci-ioda.c
> index cde710297a4e..f480e3fc9ebb 100644
> --- a/arch/powerpc/platforms/powernv/pci-ioda.c
> +++ b/arch/powerpc/platforms/powernv/pci-ioda.c
> @@ -1203,75 +1203,6 @@ static struct pnv_ioda_pe *pnv_ioda_setup_bus_PE(struct pci_bus *bus, bool all)
> return pe;
> }
>
> -static struct pnv_ioda_pe *pnv_ioda_setup_npu_PE(struct pci_dev *npu_pdev)
> -{
> - int pe_num, found_pe = false, rc;
> - long rid;
> - struct pnv_ioda_pe *pe;
> - struct pci_dev *gpu_pdev;
> - struct pci_dn *npu_pdn;
> - struct pci_controller *hose = pci_bus_to_host(npu_pdev->bus);
> - struct pnv_phb *phb = hose->private_data;
> -
> - /*
> - * Due to a hardware errata PE#0 on the NPU is reserved for
> - * error handling. This means we only have three PEs remaining
> - * which need to be assigned to four links, implying some
> - * links must share PEs.
> - *
> - * To achieve this we assign PEs such that NPUs linking the
> - * same GPU get assigned the same PE.
> - */
> - gpu_pdev = pnv_pci_get_gpu_dev(npu_pdev);
> - for (pe_num = 0; pe_num < phb->ioda.total_pe_num; pe_num++) {
> - pe = &phb->ioda.pe_array[pe_num];
> - if (!pe->pdev)
> - continue;
> -
> - if (pnv_pci_get_gpu_dev(pe->pdev) == gpu_pdev) {
> - /*
> - * This device has the same peer GPU so should
> - * be assigned the same PE as the existing
> - * peer NPU.
> - */
> - dev_info(&npu_pdev->dev,
> - "Associating to existing PE %x\n", pe_num);
> - pci_dev_get(npu_pdev);
> - npu_pdn = pci_get_pdn(npu_pdev);
> - rid = npu_pdev->bus->number << 8 | npu_pdn->devfn;
> - npu_pdn->pe_number = pe_num;
> - phb->ioda.pe_rmap[rid] = pe->pe_number;
> -
> - /* Map the PE to this link */
> - rc = opal_pci_set_pe(phb->opal_id, pe_num, rid,
> - OpalPciBusAll,
> - OPAL_COMPARE_RID_DEVICE_NUMBER,
> - OPAL_COMPARE_RID_FUNCTION_NUMBER,
> - OPAL_MAP_PE);
> - WARN_ON(rc != OPAL_SUCCESS);
> - found_pe = true;
> - break;
> - }
> - }
> -
> - if (!found_pe)
> - /*
> - * Could not find an existing PE so allocate a new
> - * one.
> - */
> - return pnv_ioda_setup_dev_PE(npu_pdev);
> - else
> - return pe;
> -}
> -
> -static void pnv_ioda_setup_npu_PEs(struct pci_bus *bus)
> -{
> - struct pci_dev *pdev;
> -
> - list_for_each_entry(pdev, &bus->devices, bus_list)
> - pnv_ioda_setup_npu_PE(pdev);
> -}
> -
> static void pnv_pci_ioda_setup_PEs(void)
> {
> struct pci_controller *hose, *tmp;
> @@ -1281,13 +1212,6 @@ static void pnv_pci_ioda_setup_PEs(void)
>
> list_for_each_entry_safe(hose, tmp, &hose_list, list_node) {
> phb = hose->private_data;
> - if (phb->type == PNV_PHB_NPU_NVLINK) {
> - /* PE#0 is needed for error reporting */
> - pnv_ioda_reserve_pe(phb, 0);
> - pnv_ioda_setup_npu_PEs(hose->bus);
> - if (phb->model == PNV_PHB_MODEL_NPU2)
> - pnv_npu2_init(phb);
> - }
> if (phb->type == PNV_PHB_NPU_OCAPI) {
> bus = hose->bus;
> list_for_each_entry(pdev, &bus->devices, bus_list)
> @@ -1895,9 +1819,6 @@ static int pnv_pci_ioda_dma_set_mask(struct pci_dev *pdev, u64 dma_mask)
> }
> *pdev->dev.dma_mask = dma_mask;
>
> - /* Update peer npu devices */
> - pnv_npu_try_dma_set_bypass(pdev, bypass);
> -
> return 0;
> }
>
> @@ -2143,14 +2064,6 @@ static void pnv_pci_ioda2_tce_invalidate(struct iommu_table *tbl,
> }
> }
>
> -void pnv_pci_ioda2_tce_invalidate_entire(struct pnv_phb *phb, bool rm)
> -{
> - if (phb->model == PNV_PHB_MODEL_NPU || phb->model == PNV_PHB_MODEL_PHB3)
> - pnv_pci_phb3_tce_invalidate_entire(phb, rm);
> - else
> - opal_pci_tce_kill(phb->opal_id, OPAL_PCI_TCE_KILL, 0, 0, 0, 0);
> -}
> -
> static int pnv_ioda2_tce_build(struct iommu_table *tbl, long index,
> long npages, unsigned long uaddr,
> enum dma_data_direction direction,
> @@ -2639,137 +2552,6 @@ static struct iommu_table_group_ops pnv_pci_ioda2_ops = {
> .take_ownership = pnv_ioda2_take_ownership,
> .release_ownership = pnv_ioda2_release_ownership,
> };
> -
> -static int gpe_table_group_to_npe_cb(struct device *dev, void *opaque)
> -{
> - struct pci_controller *hose;
> - struct pnv_phb *phb;
> - struct pnv_ioda_pe **ptmppe = opaque;
> - struct pci_dev *pdev = container_of(dev, struct pci_dev, dev);
> - struct pci_dn *pdn = pci_get_pdn(pdev);
> -
> - if (!pdn || pdn->pe_number == IODA_INVALID_PE)
> - return 0;
> -
> - hose = pci_bus_to_host(pdev->bus);
> - phb = hose->private_data;
> - if (phb->type != PNV_PHB_NPU_NVLINK)
> - return 0;
> -
> - *ptmppe = &phb->ioda.pe_array[pdn->pe_number];
> -
> - return 1;
> -}
> -
> -/*
> - * This returns PE of associated NPU.
> - * This assumes that NPU is in the same IOMMU group with GPU and there is
> - * no other PEs.
> - */
> -static struct pnv_ioda_pe *gpe_table_group_to_npe(
> - struct iommu_table_group *table_group)
> -{
> - struct pnv_ioda_pe *npe = NULL;
> - int ret = iommu_group_for_each_dev(table_group->group, &npe,
> - gpe_table_group_to_npe_cb);
> -
> - BUG_ON(!ret || !npe);
> -
> - return npe;
> -}
> -
> -static long pnv_pci_ioda2_npu_set_window(struct iommu_table_group *table_group,
> - int num, struct iommu_table *tbl)
> -{
> - struct pnv_ioda_pe *npe = gpe_table_group_to_npe(table_group);
> - int num2 = (num == 0) ? 1 : 0;
> - long ret = pnv_pci_ioda2_set_window(table_group, num, tbl);
> -
> - if (ret)
> - return ret;
> -
> - if (table_group->tables[num2])
> - pnv_npu_unset_window(npe, num2);
> -
> - ret = pnv_npu_set_window(npe, num, tbl);
> - if (ret) {
> - pnv_pci_ioda2_unset_window(table_group, num);
> - if (table_group->tables[num2])
> - pnv_npu_set_window(npe, num2,
> - table_group->tables[num2]);
> - }
> -
> - return ret;
> -}
> -
> -static long pnv_pci_ioda2_npu_unset_window(
> - struct iommu_table_group *table_group,
> - int num)
> -{
> - struct pnv_ioda_pe *npe = gpe_table_group_to_npe(table_group);
> - int num2 = (num == 0) ? 1 : 0;
> - long ret = pnv_pci_ioda2_unset_window(table_group, num);
> -
> - if (ret)
> - return ret;
> -
> - if (!npe->table_group.tables[num])
> - return 0;
> -
> - ret = pnv_npu_unset_window(npe, num);
> - if (ret)
> - return ret;
> -
> - if (table_group->tables[num2])
> - ret = pnv_npu_set_window(npe, num2, table_group->tables[num2]);
> -
> - return ret;
> -}
> -
> -static void pnv_ioda2_npu_take_ownership(struct iommu_table_group *table_group)
> -{
> - /*
> - * Detach NPU first as pnv_ioda2_take_ownership() will destroy
> - * the iommu_table if 32bit DMA is enabled.
> - */
> - pnv_npu_take_ownership(gpe_table_group_to_npe(table_group));
> - pnv_ioda2_take_ownership(table_group);
> -}
> -
> -static struct iommu_table_group_ops pnv_pci_ioda2_npu_ops = {
> - .get_table_size = pnv_pci_ioda2_get_table_size,
> - .create_table = pnv_pci_ioda2_create_table_userspace,
> - .set_window = pnv_pci_ioda2_npu_set_window,
> - .unset_window = pnv_pci_ioda2_npu_unset_window,
> - .take_ownership = pnv_ioda2_npu_take_ownership,
> - .release_ownership = pnv_ioda2_release_ownership,


These are called from the VFIO SPAPR TCE IOMMU subdriver when certain
GPUs are passed through with VFIO:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/vfio/vfio_iommu_spapr_tce.c



> -};
> -
> -static void pnv_pci_ioda_setup_iommu_api(void)
> -{
> - struct pci_controller *hose, *tmp;
> - struct pnv_phb *phb;
> - struct pnv_ioda_pe *pe, *gpe;
> -
> - /*
> - * Now we have all PHBs discovered, time to add NPU devices to
> - * the corresponding IOMMU groups.
> - */
> - list_for_each_entry_safe(hose, tmp, &hose_list, list_node) {
> - phb = hose->private_data;
> -
> - if (phb->type != PNV_PHB_NPU_NVLINK)
> - continue;
> -
> - list_for_each_entry(pe, &phb->ioda.pe_list, list) {
> - gpe = pnv_pci_npu_setup_iommu(pe);
> - if (gpe)
> - gpe->table_group.ops = &pnv_pci_ioda2_npu_ops;
> - }
> - }
> -}
> -#else /* !CONFIG_IOMMU_API */
> -static void pnv_pci_ioda_setup_iommu_api(void) { };
> #endif
>
> static unsigned long pnv_ioda_parse_tce_sizes(struct pnv_phb *phb)
> @@ -3266,7 +3048,6 @@ static void pnv_pci_enable_bridges(void)
> static void pnv_pci_ioda_fixup(void)
> {
> pnv_pci_ioda_setup_PEs();
> - pnv_pci_ioda_setup_iommu_api();
> pnv_pci_ioda_create_dbgfs();
>
> pnv_pci_enable_bridges();
> @@ -3713,27 +3494,6 @@ static const struct pci_controller_ops pnv_pci_ioda_controller_ops = {
> .shutdown = pnv_pci_ioda_shutdown,
> };
>
> -static int pnv_npu_dma_set_mask(struct pci_dev *npdev, u64 dma_mask)
> -{
> - dev_err_once(&npdev->dev,
> - "%s operation unsupported for NVLink devices\n",
> - __func__);
> - return -EPERM;
> -}
> -
> -static const struct pci_controller_ops pnv_npu_ioda_controller_ops = {
> - .dma_dev_setup = pnv_pci_dma_dev_setup,
> -#ifdef CONFIG_PCI_MSI
> - .setup_msi_irqs = pnv_setup_msi_irqs,
> - .teardown_msi_irqs = pnv_teardown_msi_irqs,
> -#endif
> - .enable_device_hook = pnv_pci_enable_device_hook,
> - .window_alignment = pnv_pci_window_alignment,
> - .reset_secondary_bus = pnv_pci_reset_secondary_bus,
> - .dma_set_mask = pnv_npu_dma_set_mask,
> - .shutdown = pnv_pci_ioda_shutdown,
> -};
> -
> static const struct pci_controller_ops pnv_npu_ocapi_ioda_controller_ops = {
> .enable_device_hook = pnv_pci_enable_device_hook,
> .window_alignment = pnv_pci_window_alignment,
> @@ -3955,9 +3715,6 @@ static void __init pnv_pci_init_ioda_phb(struct device_node *np,
> ppc_md.pcibios_fixup = pnv_pci_ioda_fixup;
>
> switch (phb->type) {
> - case PNV_PHB_NPU_NVLINK:
> - hose->controller_ops = pnv_npu_ioda_controller_ops;
> - break;
> case PNV_PHB_NPU_OCAPI:
> hose->controller_ops = pnv_npu_ocapi_ioda_controller_ops;
> break;
> diff --git a/arch/powerpc/platforms/powernv/pci.h b/arch/powerpc/platforms/powernv/pci.h
> index 8b37b28e3831..54f2935b7ac5 100644
> --- a/arch/powerpc/platforms/powernv/pci.h
> +++ b/arch/powerpc/platforms/powernv/pci.h
> @@ -231,17 +231,6 @@ extern void pe_level_printk(const struct pnv_ioda_pe *pe, const char *level,
> #define pe_info(pe, fmt, ...) \
> pe_level_printk(pe, KERN_INFO, fmt, ##__VA_ARGS__)
>
> -/* Nvlink functions */
> -extern void pnv_npu_try_dma_set_bypass(struct pci_dev *gpdev, bool bypass);
> -extern void pnv_pci_ioda2_tce_invalidate_entire(struct pnv_phb *phb, bool rm);
> -extern struct pnv_ioda_pe *pnv_pci_npu_setup_iommu(struct pnv_ioda_pe *npe);
> -extern long pnv_npu_set_window(struct pnv_ioda_pe *npe, int num,
> - struct iommu_table *tbl);
> -extern long pnv_npu_unset_window(struct pnv_ioda_pe *npe, int num);
> -extern void pnv_npu_take_ownership(struct pnv_ioda_pe *npe);
> -extern void pnv_npu_release_ownership(struct pnv_ioda_pe *npe);
> -extern int pnv_npu2_init(struct pnv_phb *phb);
> -
> /* pci-ioda-tce.c */
> #define POWERNV_IOMMU_DEFAULT_LEVELS 1
> #define POWERNV_IOMMU_MAX_LEVELS 5
>


--
Alexey