Linaro-mm-sig

linaro-mm-sig@lists.linaro.org

14 participants
3627 discussions

Re: [RFC PATCH 2/6] dma-buf: Add get_phys() to describe a physical run

by Christian König

On 10/5/26 15:20, Fred Griffoul wrote: > On 10/5/26 Christian König wrote: >> Well filling page tables by the importer is an absolutely clear NO-GO >> for the DMA-buf design, we have gotten down that path already and it >> took us years to remove this functionality again. > > Understood, thanks for the quick and clear answers on this patch and on > patch 3. > >> The problem you are facing here is that dma_buf_mmap() doesn't work >> because you don't have a VMA. > > Right: the memory has no struct page and is deliberately not mapped in > the host, so there is no VMA to hand to KVM. > > I'll post a new version of this series. It drops get_phys(), the ranged > invalidation and the guest_memfd dma-buf import (patches 2-5), and it > does not touch drivers/dma-buf, as you suggested in the PAL discussion: > > https://lore.kernel.org/all/c413710b-4c28-4ed8-88ec-aeb8c4482011@amd.com/ > > KVM gets its frames through the KVM-private guest_memfd provider > interface from David's series. iommufd gets them through a private > interface with the exporter, as it already does for vfio-pci. The new > version adds a small registry in iommufd for that, which replaces the > symbol_get() of the sample in David's series. The dma-buf is then only > the handle and the existing revoke. Yeah that approach sounds totally sane to me. We have drivers which stuff all kind of resources, not just memory, into a DMA-buf. That ranges from MMIO BARs to trigger FW operations (doorbells) over full HW blocks (ordered appends, global wave sync etc...) where two or more applications need to share which one is used. That you then have a exporter private interface is for accessing this is perfectly ok. The problem you are facing here is more that this is not limited to one driver/module but multiple ones and you don't want module inter dependencies because of the symbols. So symbol_get() is probably the right thing to do as long as we don't have something like weak symbols or similar. Regards, Christian. > > Thanks, > Fred

1 hour, 12 minutes

Re: [PATCH v9 01/13] dma-buf: introduce initial file I/O infrastructure

by Matthew Brost

On Wed, Oct 07, 2026 at 02:42:44AM +0100, Pavel Begunkov wrote: > The goal is to be able to natively use dma-buf in the read-write / IO > path. This patch adds basic building blocks serving as a glue and API > between drivers and upper layer subsystems providing the uAPI. Later > patches implement it for NVMe raw block devices and expose it to the > user space via io_uring. > > There are two main objects. struct dma_buf_io_ctx and struct > dma_buf_io_map. The ctx is used during initial registration and serves > as an interface between the upper layer user like io_uring and to the > importer subsystem / driver. The map represents the actual dma map > established for the target device[s] with dma_buf_map_attachment() and > stored in a device specific format. The context is created via a new > file operation ->init_dma_buf_io_ctx. > > The ctx-map separation exists to facilitate map invalidation (see > dma_buf_io_invalidate_mappings()). A ctx can create > multiple maps during its lifetime, but there can only be no more than > one (active) map attached to it. Invalidation drops the active map > if present, and any new I/O request will try to create a new one. > > The primary task of dma_buf_io_map is to count requests using it > and wait for their completion when we want to destroy the DMA map. > > [un]mapping and any work with dma addresses is delegated to the > importer driver via an ops table stored in the ctx, see struct > dma_buf_io_ops. Only the target driver / subsystem knows about devices > it wants to use the dma-buf with, especially in case of multi-device > filesystems or stacking in the future. > Full disclosure Pavel incorporated some of my ideas wrt to locking cleanup which I shared off the list. So I can't in fairness give an RB but: Acked-by: Matthew Brost <matthew.brost(a)intel.com> For the series - I can give this and I have pulled to code + have Xe nvme + IGT setup showing this working: Tested-by: Matthew Brost <matthew.brost(a)intel.com> > Signed-off-by: Pavel Begunkov <asml.silence(a)gmail.com> > --- > drivers/dma-buf/Makefile | 2 +- > drivers/dma-buf/dma-buf-io.c | 183 +++++++++++++++++++++++++++++++++++ > include/linux/dma-buf-io.h | 105 ++++++++++++++++++++ > include/linux/fs.h | 2 + > 4 files changed, 291 insertions(+), 1 deletion(-) > create mode 100644 drivers/dma-buf/dma-buf-io.c > create mode 100644 include/linux/dma-buf-io.h > > diff --git a/drivers/dma-buf/Makefile b/drivers/dma-buf/Makefile > index b25d7550bacf..523731b0f83e 100644 > --- a/drivers/dma-buf/Makefile > +++ b/drivers/dma-buf/Makefile > @@ -1,6 +1,6 @@ > # SPDX-License-Identifier: GPL-2.0-only > obj-y := dma-buf.o dma-fence.o dma-fence-array.o dma-fence-chain.o \ > - dma-fence-unwrap.o dma-resv.o dma-buf-mapping.o > + dma-fence-unwrap.o dma-resv.o dma-buf-mapping.o dma-buf-io.o > obj-$(CONFIG_DMABUF_HEAPS) += dma-heap.o > obj-$(CONFIG_DMABUF_HEAPS) += heaps/ > obj-$(CONFIG_SYNC_FILE) += sync_file.o > diff --git a/drivers/dma-buf/dma-buf-io.c b/drivers/dma-buf/dma-buf-io.c > new file mode 100644 > index 000000000000..9ba9c17f900a > --- /dev/null > +++ b/drivers/dma-buf/dma-buf-io.c > @@ -0,0 +1,183 @@ > +/* SPDX-License-Identifier: GPL-2.0 */ > +/* > + * Common infrastructure for supporing dma-buf in the I/O path. > + * > + * Copyright (C) 2026 Pavel Begunkov <asml.silence(a)gmail.com> > + */ > +#include <linux/dma-buf-io.h> > +#include <linux/dma-resv.h> > + > +static void dma_buf_io_map_release(struct dma_buf_io_map *map) > +{ > + struct dma_buf_io_ctx *ctx = map->ctx; > + > + dma_resv_assert_held(ctx->dmabuf->resv); > + > + percpu_ref_exit(&map->refs); > + ctx->dev_ops->unmap(ctx, map); > +} > + > +static void dma_buf_io_map_refs_cb(struct percpu_ref *ref) > +{ > + struct dma_buf_io_map *map = container_of(ref, struct dma_buf_io_map, refs); > + > + complete(&map->drained); > +} > + > +static void dma_buf_io_kill_map(struct dma_buf_io_ctx *ctx) > +{ > + struct dma_buf_io_map *map; > + > + dma_resv_assert_held(ctx->dmabuf->resv); > + > + map = rcu_dereference_protected(ctx->map, > + dma_resv_held(ctx->dmabuf->resv)); > + if (!map) > + return; > + > + rcu_assign_pointer(ctx->map, NULL); > + percpu_ref_kill(&map->refs); > + /* make sure the map is not visible via ctx->map */ > + synchronize_rcu_expedited(); > + wait_for_completion(&map->drained); > + dma_buf_io_map_release(map); > +} > + > +int dma_buf_io_init_map(struct dma_buf_io_ctx *ctx, struct dma_buf_io_map *map, > + struct sg_table *sgt) > +{ > + unsigned seg_shift = ~0U; > + struct scatterlist *sg; > + unsigned long tmp; > + int ret; > + > + for_each_sgtable_dma_sg(sgt, sg, tmp) > + seg_shift = min(seg_shift, __ffs(sg_dma_len(sg))); > + > + ret = percpu_ref_init(&map->refs, dma_buf_io_map_refs_cb, 0, GFP_KERNEL); > + if (ret) > + return ret; > + init_completion(&map->drained); > + map->min_seg_shift = seg_shift; > + map->ctx = ctx; > + return 0; > +} > +EXPORT_SYMBOL_NS_GPL(dma_buf_io_init_map, "DMA_BUF"); > + > +static struct dma_buf_io_map *__dma_buf_io_create_map(struct dma_buf_io_ctx *ctx) > +{ > + struct dma_buf *dmabuf = ctx->dmabuf; > + struct dma_buf_io_map *map; > + long ret; > + > + dma_resv_assert_held(dmabuf->resv); > + > + if (ctx->is_dead) > + return ERR_PTR(-ENOENT); > + > + map = __dma_buf_io_get_map(ctx); > + if (map) > + return map; > + > + map = ctx->dev_ops->map(ctx); > + if (IS_ERR(map)) > + return map; > + > + if (WARN_ON_ONCE(!map->min_seg_shift)) > + return ERR_PTR(-EFAULT); > + > + ret = dma_resv_wait_timeout(dmabuf->resv, DMA_RESV_USAGE_KERNEL, > + true, MAX_SCHEDULE_TIMEOUT); > + if (ret <= 0) { > + if (!ret) > + ret = -EAGAIN; > + dma_buf_io_map_release(map); > + return ERR_PTR(ret); > + } > + > + /* get a reference for the caller */ > + percpu_ref_get(&map->refs); > + rcu_assign_pointer(ctx->map, map); > + return map; > +} > + > +struct dma_buf_io_map *dma_buf_io_create_map(struct dma_buf_io_ctx *ctx) > +{ > + struct dma_buf_io_map *map; > + int ret; > + > + ret = dma_resv_lock_interruptible(ctx->dmabuf->resv, NULL); > + if (ret) > + return ERR_PTR(ret); > + > + map = __dma_buf_io_create_map(ctx); > + dma_resv_unlock(ctx->dmabuf->resv); > + return map; > +} > + > +static void dma_buf_io_ctx_shutdown(struct dma_buf_io_ctx *ctx) > +{ > + dma_resv_lock(ctx->dmabuf->resv, NULL); > + ctx->is_dead = true; > + dma_buf_io_kill_map(ctx); > + dma_resv_unlock(ctx->dmabuf->resv); > +} > + > +int dma_buf_io_ctx_create(struct file *file, > + struct dma_buf *dmabuf, > + enum dma_data_direction dir, > + struct dma_buf_io_ctx **out_ctx) > +{ > + struct dma_buf_io_ctx *ctx; > + int ret; > + > + if (!file->f_op->init_dma_buf_io_ctx) > + return -EOPNOTSUPP; > + > + ctx = kzalloc_obj(*ctx); > + if (!ctx) > + return -ENOMEM; > + ctx->dir = dir; > + ctx->dmabuf = dmabuf; > + get_dma_buf(dmabuf); > + > + ret = file->f_op->init_dma_buf_io_ctx(file, ctx); > + if (ret) { > + kfree(ctx); > + dma_buf_put(dmabuf); > + return ret; > + } > + > + if (WARN_ON_ONCE(!ctx->dev_ops || > + !ctx->dev_ops->map || > + !ctx->dev_ops->unmap || > + !ctx->dev_ops->release)) > + return -EINVAL; > + > + *out_ctx = ctx; > + return 0; > +} > + > +void dma_buf_io_ctx_release(struct dma_buf_io_ctx *ctx) > +{ > + dma_buf_io_ctx_shutdown(ctx); > + > + if (WARN_ON_ONCE(rcu_dereference_protected(ctx->map, true))) > + return; > + > + ctx->dev_ops->release(ctx); > + dma_buf_put(ctx->dmabuf); > + kfree(ctx); > +} > + > +void dma_buf_io_invalidate_mappings(struct dma_buf_io_ctx *ctx) > +{ > + dma_buf_io_kill_map(ctx); > +} > +EXPORT_SYMBOL_NS_GPL(dma_buf_io_invalidate_mappings, "DMA_BUF"); > + > +void dma_buf_io_detach(struct dma_buf_io_ctx *ctx) > +{ > + dma_buf_io_ctx_shutdown(ctx); > +} > +EXPORT_SYMBOL_NS_GPL(dma_buf_io_detach, "DMA_BUF"); > diff --git a/include/linux/dma-buf-io.h b/include/linux/dma-buf-io.h > new file mode 100644 > index 000000000000..8ac338df51ff > --- /dev/null > +++ b/include/linux/dma-buf-io.h > @@ -0,0 +1,105 @@ > +/* SPDX-License-Identifier: GPL-2.0 */ > +#ifndef __DMA_BUF_IO_H__ > +#define __DMA_BUF_IO_H__ > + > +#include <linux/dma-buf.h> > + > +struct dma_buf_io_ctx; > +struct dma_buf_io_map; > + > +struct dma_buf_io_ops { > + /* > + * Create a new map for the given ctx. Called with the reservation > + * lock held. > + */ > + struct dma_buf_io_map *(*map)(struct dma_buf_io_ctx *ctx); > + > + /* > + * Clean up device specific parts of the @map. Called with the > + * reservation lock held. > + */ > + void (*unmap)(struct dma_buf_io_ctx *ctx, struct dma_buf_io_map *map); > + > + /* > + * The user tries to destroy the ctx. Release all device specific > + * parts of the token. > + */ > + void (*release)(struct dma_buf_io_ctx *); > +}; > + > +struct dma_buf_io_map { > + /* > + * Counts attached requests and other users. Device specific unmapping > + * is deferred until all refs are dropped. > + */ > + struct percpu_ref refs; > + /* > + * Shift for the minimum segment size of the mapping. > + */ > + unsigned min_seg_shift; > + > + struct dma_buf_io_ctx *ctx; > + struct completion drained; > +}; > + > +struct dma_buf_io_ctx { > + struct dma_buf_io_map __rcu *map; > + struct dma_buf *dmabuf; > + enum dma_data_direction dir; > + bool is_dead; > + > + void *dev_priv; > + const struct dma_buf_io_ops *dev_ops; > +}; > + > +int dma_buf_io_ctx_create(struct file *file, > + struct dma_buf *dmabuf, > + enum dma_data_direction dir, > + struct dma_buf_io_ctx **ctx); > +void dma_buf_io_ctx_release(struct dma_buf_io_ctx *ctx); > + > +struct dma_buf_io_map *dma_buf_io_create_map(struct dma_buf_io_ctx *ctx); > + > +static inline struct dma_buf_io_map * > +__dma_buf_io_get_map(struct dma_buf_io_ctx *ctx) > +{ > + struct dma_buf_io_map *map; > + > + guard(rcu)(); > + > + map = rcu_dereference(ctx->map); > + if (unlikely(!map || !percpu_ref_tryget_live_rcu(&map->refs))) > + return NULL; > + > + return map; > +} > + > +static inline struct dma_buf_io_map * > +dma_buf_io_get_map(struct dma_buf_io_ctx *ctx, bool nowait) > +{ > + struct dma_buf_io_map *map; > + > + map = __dma_buf_io_get_map(ctx); > + if (likely(map)) > + return map; > + > + if (nowait) > + return ERR_PTR(-EAGAIN); > + return dma_buf_io_create_map(ctx); > +} > + > +static inline void dma_buf_io_map_drop(struct dma_buf_io_map *map) > +{ > + percpu_ref_put(&map->refs); > +} > + > +/* > + * Device API > + */ > + > +void dma_buf_io_invalidate_mappings(struct dma_buf_io_ctx *ctx); > +int dma_buf_io_init_map(struct dma_buf_io_ctx *ctx, struct dma_buf_io_map *map, > + struct sg_table *sgt); > +void dma_buf_io_detach(struct dma_buf_io_ctx *ctx); > + > +#endif > diff --git a/include/linux/fs.h b/include/linux/fs.h > index f9d1e05e8ae6..05c1ff495732 100644 > --- a/include/linux/fs.h > +++ b/include/linux/fs.h > @@ -1914,6 +1914,7 @@ struct dir_context { > #define COPY_FILE_SPLICE (1 << 0) > > struct io_uring_cmd; > +struct dma_buf_io_ctx; > struct offset_ctx; > > struct file_operations { > @@ -1960,6 +1961,7 @@ struct file_operations { > int (*uring_cmd_iopoll)(struct io_uring_cmd *, struct io_comp_batch *, > unsigned int poll_flags); > int (*mmap_prepare)(struct vm_area_desc *); > + int (*init_dma_buf_io_ctx)(struct file *, struct dma_buf_io_ctx *); > } __randomize_layout; > > /* Supports async buffered reads */ > -- > 2.54.0 >

2 hours, 50 minutes

[PATCH v10 00/18] PCI/P2PDMA: Route peer-to-peer DMA by TLP class

by Leon Romanovsky

PCI P2PDMA applies Request and Completion Redirect throughout both paths. This misclassifies asymmetric and nested switches, and reports one answer for every kind of TLP. Three ACS controls act on TLP attributes the Requester chooses rather than on the topology: Translation Blocking and Direct Translated P2P act on a Request's Address Type, and Completion Redirect skips Completions carrying Relaxed Ordering. Evaluate each direction at the path divergence, decide every class from the one walk, and treat a client with ATS enabled as translating unless its driver declares per-mapping ATS. This completes the P2PDMA side; dma-buf and mlx5 follow separately. Signed-off-by: Leon Romanovsky <leonro(a)nvidia.com> --- Changes in v10: - Fixed the Relaxed Ordering description: a Completion carries the attribute of its Request (PCIe r7.0 sec 2.2.9.1). - Used the PCIe terms Requester and Completer for TLPs, ACS and ports; client and provider remain for P2PDMA objects. - Rebased on pci/p2pdma, which includes the prerequisite P2PDMA fixes. - Added a topology example to the path divergence patch. - Removed a repeated sentence from the disable_acs_redir documentation. - Explained the path divergence in the documentation. - Link to v9: https://patch.msgid.link/20261001-fix-p2p-acs-v4-0-v9-0-1a8e0f50ddd9@nvidia… Changes in v9: - Removed dmabuf patches for now. - Removed mlx5 per-mapping ATS declaration for now; it comes back with the dmabuf patches. - Link to v8: https://patch.msgid.link/20260928-fix-p2p-acs-v4-0-v8-0-404453b9c435@nvidia… Changes in v8: - Added extra Reviewed-by from Logan. - Added patches to take care of ATS per-device vs. per-mapping option - Link to v7: https://patch.msgid.link/20260920-fix-p2p-acs-v4-0-v7-0-ca0828ab697c@nvidia… Changes in v7: - Removed "Document pdev->p2pdma lifetime rules" patch, it gives nothing after p2pmem fix. - Split dmabuf patch. - Tushar retested the series, so added his Tested-by. - Added Logan's ROB tags and fixed minor documentation issues pointed by him. - Link to v6: https://lore.kernel.org/all/20260914-fix-p2p-acs-v4-0-v6-0-5ef07ec9ef06@nvi… Changes in v6: - Changed DMABUF to use callbacks and not directly stored pointer. - Link to v5: https://patch.msgid.link/20260910-fix-p2p-acs-v4-0-v5-0-856087f63c0d@nvidia… Changes in v5: - Rebase on the posted fixes. - Dropped tags from changed patches. - Remove Egress Control Vector interpretation and coverage. - Keep enabled Egress Control conservative as a Request redirect. - Use pci_dbg()/dev_dbg() for diagnostics and drop the "debug" prefix. - Removed code comments from "Document the pdev->p2pdma lifetime and RCU rules" patch and reduced description to actual lifetime explanation. - Added note that Linux assumes that TLPs are in strict-ordering and untranslated. - Added code to calculate p2p paths per-TLP type. - Converted mlx5 to use that new proposed API. - Link to https://patch.msgid.link/20260821-fix-p2p-acs-v4-0-v4-0-94426b96de73@nvidia… Changes in v4: - Reject ACS Violations and unreadable routing state instead of treating them as host-bridge redirects - Added Tested-by tags from Tushar Dave - Added support to asymmetric ACS routing - Limited redirect checks to the two ports at the path divergence - Added standalone ACS routing diagnostics for hardware retesting - Dropped " PCI: Account for Direct Translated P2P in ACS isolation checks" patch - Link to v3: https://patch.msgid.link/20260811-fix-p2p-acs-v3-0-efc488ee7c03@nvidia.com Changes in v3: - Fixed pci_p2pdma_add_resource() error unwinding - Made pdev->p2pdma teardown wait unconditionally for RCU readers - Restricted pci_p2pmem_find_many() to pool-backed providers - Documented the pdev->p2pdma lifetime and RCU rules - Fixed calc_map_type_and_dist() handling of the verbose argument - Required the ACS port and target to share a bus before indexing the Egress Control Vector - Gave pci_acs_enabled() and pci_acs_path_enabled() a scope, so the ACS Direct Translated P2P rule no longer stops pci_enable_pasid() from enabling PASID - Dropped "Report ACS ports when the paths share no upstream bridge": the mapping type cannot change without a shared upstream bridge, so the pci=disable_acs_redir= hint was not actionable there and the ACS walk only cost config space reads - Folded the Request Redirect rule into pci_acs_rr_ineffective(), so pci_acs_flags_enabled() and the Intel SPT PCH quirk share one copy - Renamed pci_acs_egress_ctrl_set() to pci_acs_egress_ctrl_is_set(), it reads the bit rather than setting it - Reworded the blocked-path warning: ACS may also leave the direct route indeterminate rather than blocked - Added KUnit coverage for the shared-bus guard, a device with no ACS capability and an unreadable ACS Control register - Added the missing Fixes: tags, a second one on the pci_p2pdma_add_resource() unwinding fix (the dangling devres action dates to f58ef9d1d135) and one on the Egress Control isolation change - Link to v2: https://patch.msgid.link/20260806-fix-p2p-acs-v2-0-0cec14812965@nvidia.com Changes in v2: - Added Logan's ROB tags - Added commas in Documentation patch - Link to v1: https://patch.msgid.link/20260802-fix-p2p-acs-v1-0-a7c5eb64fff6@nvidia.com --- Leon Romanovsky (18): PCI/P2PDMA: Document the TLP attribute assumptions PCI/P2PDMA: Derive routing from directional ACS controls PCI: Reject unreadable ACS controls in isolation checks PCI/P2PDMA: Evaluate ACS controls at the path divergence PCI/P2PDMA: Document directional ACS routing PCI/P2PDMA: Collect the path's ACS controls before deciding PCI/P2PDMA: Answer routing per TLP class PCI/P2PDMA: Route Relaxed Ordering Completions directly PCI/P2PDMA: Reject Translated Requests blocked by Translation Blocking PCI/P2PDMA: Route Translated Requests under Direct Translated P2P PCI/P2PDMA: Log detailed ACS routing diagnostics PCI/P2PDMA: Add KUnit tests for the ACS routing decisions PCI/P2PDMA: Test the ACS P2P routing walk PCI: Add KUnit coverage for ACS isolation checks PCI/P2PDMA: Document TLP-class routing PCI/P2PDMA: Let a client declare that it selects ATS per mapping PCI/P2PDMA: Evaluate the ATS path for clients with ATS enabled PCI/P2PDMA: Test the routing of clients with ATS enabled Documentation/admin-guide/kernel-parameters.txt | 11 +- Documentation/driver-api/pci/p2pdma.rst | 81 +++ drivers/pci/Kconfig | 15 + drivers/pci/Makefile | 1 + drivers/pci/p2pdma.c | 719 ++++++++++++++++++-- drivers/pci/pci.c | 7 +- drivers/pci/pci.h | 56 ++ drivers/pci/pci_acs_test.c | 860 ++++++++++++++++++++++++ drivers/pci/quirks.c | 6 +- include/linux/pci-p2pdma.h | 12 +- 10 files changed, 1683 insertions(+), 85 deletions(-) --- base-commit: 36038f0e8377ef316b723d0d85b1510bc64069ab change-id: 20260821-fix-p2p-acs-v4-0-e72455e3a261 Best regards, -- Leon Romanovsky <leonro(a)nvidia.com>

2 hours, 51 minutes

[PATCH 0/2] drm/syncobj: fix sync_file import into timeline points

by Julian Orth

Importing a sync_file into a timeline point (DRM_SYNCOBJ_FD_TO_HANDLE_FLAGS_IMPORT_SYNC_FILE | DRM_SYNCOBJ_FD_TO_HANDLE_FLAGS_TIMELINE) has two bugs since c2d3a7300695: 1. The fence and syncobj references leak if dma_fence_chain_alloc() fails. 2. A sync_file that contains a dma_fence_chain (e.g. one exported from a timeline point) is wrapped in another chain and triggers the WARN_ON in dma_fence_chain_init(). I reported this on 2026-05-20. These patches were developed completely autonomously by AI. I have reviewed them, but they have not been tested at runtime. They are only compile-tested. Julian Orth (2): drm/syncobj: fix fence and syncobj leak in drm_syncobj_import_sync_file_fence drm/syncobj: flatten chains when importing sync_file into timeline point drivers/gpu/drm/drm_syncobj.c | 20 +++++++++++++++++--- 1 file changed, 17 insertions(+), 3 deletions(-) base-commit: 6ccf996a0dec1852dab94ad865f27b4904750e12 -- 2.56.0

2 hours, 52 minutes

[PATCH v9 00/18] PCI/P2PDMA: Route peer-to-peer DMA by TLP class

by Leon Romanovsky

PCI P2PDMA applies Request and Completion Redirect throughout both paths. This misclassifies asymmetric and nested switches, and reports one answer for every kind of TLP. Three ACS controls act on TLP attributes the client chooses rather than on the topology: Translation Blocking and Direct Translated P2P act on a Request's Address Type, and Completion Redirect skips Completions carrying Relaxed Ordering. Evaluate each direction at the path divergence, decide every class from the one walk, and treat a client with ATS enabled as translating unless its driver declares per-mapping ATS. This completes the P2PDMA side; dma-buf and mlx5 follow separately. Signed-off-by: Leon Romanovsky <leonro(a)nvidia.com> --- Changes in v9: - Removed dmabuf patches for now. - Removed mlx5 per-mapping ATS declaration for now; it comes back with the dmabuf patches. - Link to v8: https://patch.msgid.link/20260928-fix-p2p-acs-v4-0-v8-0-404453b9c435@nvidia… Changes in v8: - Added extra Reviewed-by from Logan. - Added patches to take care of ATS per-device vs. per-mapping option - Link to v7: https://patch.msgid.link/20260920-fix-p2p-acs-v4-0-v7-0-ca0828ab697c@nvidia… Changes in v7: - Removed "Document pdev->p2pdma lifetime rules" patch, it gives nothing after p2pmem fix. - Split dmabuf patch. - Tushar retested the series, so added his Tested-by. - Added Logan's ROB tags and fixed minor documentation issues pointed by him. - Link to v6: https://lore.kernel.org/all/20260914-fix-p2p-acs-v4-0-v6-0-5ef07ec9ef06@nvi… Changes in v6: - Changed DMABUF to use callbacks and not directly stored pointer. - Link to v5: https://patch.msgid.link/20260910-fix-p2p-acs-v4-0-v5-0-856087f63c0d@nvidia… Changes in v5: - Rebase on the posted fixes. - Dropped tags from changed patches. - Remove Egress Control Vector interpretation and coverage. - Keep enabled Egress Control conservative as a Request redirect. - Use pci_dbg()/dev_dbg() for diagnostics and drop the "debug" prefix. - Removed code comments from "Document the pdev->p2pdma lifetime and RCU rules" patch and reduced description to actual lifetime explanation. - Added note that Linux assumes that TLPs are in strict-ordering and untranslated. - Added code to calculate p2p paths per-TLP type. - Converted mlx5 to use that new proposed API. - Link to https://patch.msgid.link/20260821-fix-p2p-acs-v4-0-v4-0-94426b96de73@nvidia… Changes in v4: - Reject ACS Violations and unreadable routing state instead of treating them as host-bridge redirects - Added Tested-by tags from Tushar Dave - Added support to asymmetric ACS routing - Limited redirect checks to the two ports at the path divergence - Added standalone ACS routing diagnostics for hardware retesting - Dropped " PCI: Account for Direct Translated P2P in ACS isolation checks" patch - Link to v3: https://patch.msgid.link/20260811-fix-p2p-acs-v3-0-efc488ee7c03@nvidia.com Changes in v3: - Fixed pci_p2pdma_add_resource() error unwinding - Made pdev->p2pdma teardown wait unconditionally for RCU readers - Restricted pci_p2pmem_find_many() to pool-backed providers - Documented the pdev->p2pdma lifetime and RCU rules - Fixed calc_map_type_and_dist() handling of the verbose argument - Required the ACS port and target to share a bus before indexing the Egress Control Vector - Gave pci_acs_enabled() and pci_acs_path_enabled() a scope, so the ACS Direct Translated P2P rule no longer stops pci_enable_pasid() from enabling PASID - Dropped "Report ACS ports when the paths share no upstream bridge": the mapping type cannot change without a shared upstream bridge, so the pci=disable_acs_redir= hint was not actionable there and the ACS walk only cost config space reads - Folded the Request Redirect rule into pci_acs_rr_ineffective(), so pci_acs_flags_enabled() and the Intel SPT PCH quirk share one copy - Renamed pci_acs_egress_ctrl_set() to pci_acs_egress_ctrl_is_set(), it reads the bit rather than setting it - Reworded the blocked-path warning: ACS may also leave the direct route indeterminate rather than blocked - Added KUnit coverage for the shared-bus guard, a device with no ACS capability and an unreadable ACS Control register - Added the missing Fixes: tags, a second one on the pci_p2pdma_add_resource() unwinding fix (the dangling devres action dates to f58ef9d1d135) and one on the Egress Control isolation change - Link to v2: https://patch.msgid.link/20260806-fix-p2p-acs-v2-0-0cec14812965@nvidia.com Changes in v2: - Added Logan's ROB tags - Added commas in Documentation patch - Link to v1: https://patch.msgid.link/20260802-fix-p2p-acs-v1-0-a7c5eb64fff6@nvidia.com --- Leon Romanovsky (18): PCI/P2PDMA: Document the TLP attribute assumptions PCI/P2PDMA: Derive routing from directional ACS controls PCI: Reject unreadable ACS controls in isolation checks PCI/P2PDMA: Evaluate ACS controls at the path divergence PCI/P2PDMA: Document directional ACS routing PCI/P2PDMA: Collect the path's ACS controls before deciding PCI/P2PDMA: Answer routing per TLP class PCI/P2PDMA: Route Relaxed Ordering Completions directly PCI/P2PDMA: Reject Translated Requests blocked by Translation Blocking PCI/P2PDMA: Route Translated Requests under Direct Translated P2P PCI/P2PDMA: Log detailed ACS routing diagnostics PCI/P2PDMA: Add KUnit tests for the ACS routing decisions PCI/P2PDMA: Test the ACS P2P routing walk PCI: Add KUnit coverage for ACS isolation checks PCI/P2PDMA: Document TLP-class routing PCI/P2PDMA: Let a client declare that it selects ATS per mapping PCI/P2PDMA: Evaluate the ATS path for clients with ATS enabled PCI/P2PDMA: Test the routing of clients with ATS enabled Documentation/admin-guide/kernel-parameters.txt | 15 +- Documentation/driver-api/pci/p2pdma.rst | 80 +++ drivers/pci/Kconfig | 15 + drivers/pci/Makefile | 1 + drivers/pci/p2pdma.c | 719 ++++++++++++++++++-- drivers/pci/pci.c | 7 +- drivers/pci/pci.h | 58 ++ drivers/pci/pci_acs_test.c | 860 ++++++++++++++++++++++++ drivers/pci/quirks.c | 6 +- include/linux/pci-p2pdma.h | 12 +- 10 files changed, 1687 insertions(+), 86 deletions(-) --- base-commit: 08dbfad3f5040f5bdb6c529da20d6d4e81fefd72 change-id: 20260821-fix-p2p-acs-v4-0-e72455e3a261 prerequisite-message-id: <20260830-batch-p2p-fixes-v1-0-5044e8dfbe2e(a)nvidia.com> prerequisite-patch-id: 6b25c7fcf164cdfc14e9fac5b908d97fcf6509d7 prerequisite-patch-id: 0d083c281001365aae4b35544cf28891a6ab9a96 prerequisite-patch-id: bfd9dabf271f3cc9a3a61f46387d20c20311363d prerequisite-patch-id: fad0275efc722830fc591509506c0a5e4f581073 prerequisite-patch-id: 0c83bee688fec1f6d1564654df7c630fa6a4a978 Best regards, -- Leon Romanovsky <leonro(a)nvidia.com>

4 hours, 1 minute

[PATCH v8 0/9] vfio/pci: Add mmap() for DMABUFs

by Matt Evans

Hi all, The goal of this series is to enable userspace driver designs that use VFIO to export DMABUFs representing subsets of PCI device BARs, and "vend" those buffers from a primary process to other subordinate processes by fd. This is achieved by allowing the processes to mmap() the DMABUFs; their access to the device is isolated to the exported ranges. The primary userspace driver process can forcibly revoke access to previously-shared buffers upon cleanup (without requiring cooperation from the subordinate processes). This is an improvement on sharing the VFIO device fd to subordinate processes, which would allow unfettered access. See the RFCs for background. The existing VFIO PCI BAR mmap() becomes backed by a DMABUF too, keeping common vm_ops and fault handler for VMAs from the VFIO device and explicitly-exported DMABUFs. This will help future iommufd emulation of VFIO Type1 peer-to-peer, making it easier to get a DMABUF for a VFIO BAR as a DMA target. Below are per-patch notes & background info, and at the bottom are several related questions that reviewers may like to consider (worth at least skipping to #1, a possible bug). Notes on patches ================ mmap() conversion to use DMABUF underneath has been done for vfio-pci, but not sub-drivers: nvgrace-gpu's mmap() override path is unchanged; I kept this out of scope for now not least because I don't have a thorough test setup for this system. I would prefer to help the nvgrace-gpu maintainers enable BAR mmap() DMABUFs themselves. It's also worth noting that one reason for nvgrace-gpu's special BAR handling is to map RESMEM with a WC memory type. This is unchanged for existing mmap() but WC is not currently supported for mapping DMABUFs exported for that BAR; all DMABUF VMAs currently have the same memory type and control of UC/WC for a mapping is future work. vfio/pci: Remove DMABUF export dependency on vdev->memory_lock In v5 of this series we discover that doing an export from the VFIO mmap() path (fundamental!) adds a dependency between mmap_lock and vdev->memory_lock(W) because export was using memory_lock(W) to protect the vdev->dmabufs list and state within, and a deadlock scenario leapt out to bite: https://lore.kernel.org/kvm/9a615f22-c0d4-46ae-9654-db11e94e5fec@ozlabs.org/ The suggestion was to add a dedicated mutex/rwsem specifically for the DMABUFs/list, which is cleaner than overloading memory_lock(W). But whilst export could now downgrade to holding memory_lock(R) to test __vfio_pci_memory_enabled(), a very similar deadlock can still arise due to a memory_lock(W) elsewhere depending on a prior memory_lock(R) to be released, and attempting to take memory_lock(R) will queue behind the (W) for fairness (effectively an R->R dependency). I'd overlooked that rwsem cannot guarantee multiple readers. To be able to export while holding mmap_lock, export cannot hold memory_lock at all. Instead of testing __vfio_pci_memory_enabled() (which tracks PCI_COMMAND.MSE and PM state), this patch tracks device-global DMABUF revocation state in a new vdev->bars_revoked flag updated by vfio_pci_dma_buf_move(). If an mmap() is somehow performed during a period when DMABUFs are all revoked, then the DMABUF is created revoked. Move() already bookends reset, PM transitions etc., so subsequent revocation state changes work as-is. This flag is also protected by the dmabuf_lock and thus memory_lock is not required to export. vfio/pci: Un-revoke DMABUFs in LOW_POWER_ENTRY_WITH_WAKEUP resume On LOW_POWER entry, DMABUFs have a move(revoke=true), but the runtime resume path didn't un-revoke. This adds a corresponding move(revoke=false), which will later turn into vfio_pci_unrevoke_bars(). NOTE: the unrevoke is reordered _before_ the eventfd_signal() in vfio_pci_core_runtime_resume() to remove a window in which a waking waiter could have observed the DMABUF state as still revoked (or pm_runtime_engaged = true). (The UAPI docs state the event means the resume's complete, so observing otherwise seemed unintended.) Also, when DMABUFs are later mmap()ed, a waiter waking and taking a fault could have seen the unrevoked state and SIGBUS just before taking the memory_lock. (If the handler gets to acquire the lock, though, the resume sequence is complete, and the handler observes pm_runtime_engaged = false.) This fix is in this series because the issue will impact CPU access to the VMA as well (once they use DMABUFs), and so it's a strict dependency of later commits. dma-buf: Provide dma_buf_set_name() Makes dma_buf_set_name() available for use (by helper patch), taking a kernel-allocated string. The pre-existing local helper becomes a wrapper copying a __user string for the set-name ioctl. vfio/pci: Add a helper to look up PFNs for DMABUFs Adds a DMABUF VMA fault handler helper to determine arbitrary-sized PFNs from ranges in DMABUF. vfio/pci: Add a helper to create a DMABUF for a BAR-map VMA Refactors DMABUF export for use by the existing export feature, and adds a helper that creates a DMABUF corresponding to a VFIO BAR mmap() request. There was a request for decent debug naming in /proc/<pid>/maps etc. comparable to the existing VFIO names: since the VMAs are DMABUFs, they have a "dmabuf:" prefix and can't be 100% identical to before. This is a user-visible change, but this patch at least now gives us extra info on the BDF & BAR being mmap()ed. The name is installed with the new dma_buf_set_name() above. vfio/pci: Convert BAR mmap() to use a DMABUF The vfio-pci core mmap() creates a DMABUF with the helper above, and the vm_ops fault handler uses the other helper to resolve the fault. Because this depends on DMABUF structs/code, CONFIG_VFIO_PCI_CORE needs to depend on CONFIG_DMA_SHARED_BUFFER. The CONFIG_VFIO_PCI_DMABUF still conditionally enables the export support code. NOTE: The user mmap()s a device fd, but the resulting VMA's vm_file becomes that of the DMABUF. The DMABUF takes ownership of the device file and put()s it on release, which maintains the existing behaviour of a VMA keeping the VFIO device open. BAR zapping then happens via the existing vfio_pci_dma_buf_move() path, which now needs to unmap PTEs in the DMABUF's address_space. NOTE: As local LLM reviews did, Sashiko might falsely worry about the DMABUF fd being obtained through /proc/pid/map_files and remapped, but the file is an anon inode and doesn't support the open op (-ENXIO) so AFAICT this is currently impossible. NOTE: A side-effect of this is that is_mergeable_vma() will be false between adjacent mappings of VFIO BARs; merging would require rebuilding a larger DMABUF representing the union, not just plugging the VMAs together. The DMABUF backing the BAR is an implicit/internal export, even when the CONFIG_VFIO_PCI_DMABUF feature is not included because CONFIG_PCI_P2PDMA is not available. Without P2PDMA, it's acceptable for the DMABUF to not have a P2PDMA provider. In this configuration, the VFIO DMABUF .attach prohibits any import, which avoids getting as far as a WARN in dma_buf_map_attachment(), which would fail without P2PDMA anyway. vfio/pci: Clean up BAR zap and revocation In general (see NOTE!) the vfio_pci_zap_bars() is now obsolete, since it unmaps PTEs in the VFIO device address_space which is now unused. This consolidates all calls (e.g. around reset) with the neighbouring vfio_pci_dma_buf_move()s into new functions, to revoke/unrevoke (making the steps clearer). NOTE: Because drivers can use their own vm_ops and override .mmap, the core must conservatively assume an overridden .mmap might still add PTEs to the VFIO device address_space and therefore still does the zap. A new flag, zap_bars_on_revoke, enables the zap when .mmap is overridden. A driver that does not need the zap can clear this to opt-out, e.g. if the driver calls down to the common mmap (and so uses DMABUFs). hisi-acc-vfio-pci does just this, and thus sets the opt-out flag. vfio/pci: Support mmap() of a VFIO DMABUF Adds mmap() for a DMABUF fd exported from vfio-pci. It was a goal to keep the VFIO device fd lifetime behaviour unchanged with respect to the DMABUFs. An application can close all device fds, and this will revoke/clean up all DMABUFs; then, no mappings or other access can be performed. When enabling mmap() of the DMABUFs, this means access through the VMA is also revoked. This complicates the fault handler because whilst the DMABUF exists, it has no guarantee that the corresponding VFIO device is still alive. Adds synchronisation ensuring the vdev is available before the locks in vdev are touched; this holds the device registration so that even if the buffer has been cleaned up, vdev hasn't been freed and so the locks can be safely taken. vfio/pci: Revoke a DMABUF on request from userspace This is mostly a rename of `revoked` to an enum, `status`, and adding a third state for a buffer: usable, revoked temporary, revoked permanent. A new VFIO feature is added, VFIO_DEVICE_FEATURE_DMA_BUF_REVOKE, which takes a DMABUF (exported from the same device) and permanently revokes it. Thus a userspace driver can guarantee any downstream consumers of a shared fd are prevented from accessing a BAR range, and that range can be reused. NOTE: This might block userspace, waiting on importers to detach. The code doing revocation in vfio_pci_dma_buf_move() is moved, to a common function used by ..._move() and this new feature. Testing ======= (The [RFC ONLY] userspace test program, which drives a QEMU bochs-display function, can be found in the GitHub branch below. It at least illustrates how the export, map, revoke, and close semantics interoperate. WIP on a follow-up with a proper vfio-selftests style test based on this -- this won't be part of this series.) This code has been tested in mapping DMABUFs of single/multiple ranges from multiple BARs, aliasing mmap()s, aliasing ranges across DMABUFs, vm_pgoff > 0, revocation, shutdown/cleanup scenarios, and hugepage mappings. No regressions observed on the VFIO selftests, or on our internal vfio-pci applications. VFIO on i386 has been build-tested. Thanks to Alex Mastro for building a (WIP) testcase for the prior mmap_lock->memory_lock issue. END === This is based on v7.3-rc6 These commits are on GitHub for easier browsing, along with "[RFC ONLY] selftests: vfio: Add standalone vfio_dmabuf_mmap_test": https://github.com/evansm7/linux/compare/v7.3-rc6...dev/mev/vfio-dmabuf-mma… Also please see... https://lore.kernel.org/all/f8efaabd-c06e-4f04-8cfa-489c148e37ba@ozlabs.org/ ("[PATCH] dma-buf: Annul dmabuf->file on file release") ...which fixes a stale file pointer issue that could affect VFIO's DMABUF export. Although the issue exists prior to this series, the addition of another dmabuf->file dereference adds additional unpleasant failure modes (for example, unmap_mapping_range() the wrong file, or good old random data dereference, etc.) Thanks for reading, Matt ================================================================================ Changelog: v8: - Rebased, on v7.3-rc6 - "dma-buf: Provide dma_buf_set_name()": Comment and "DOC: locking convention" updates. dma_buf_set_name() is is listed in the "do not hold resv" category v7: https://lore.kernel.org/all/20260924152159.49702-1-matt@ozlabs.org/ v6: https://lore.kernel.org/all/20260911214200.33793-1-matt@ozlabs.org/ v5: https://lore.kernel.org/all/20260715174737.15287-1-matt@ozlabs.org/ v4: https://lore.kernel.org/all/20260701171245.90111-1-matt@ozlabs.org/ v3: https://lore.kernel.org/all/20260610154327.37758-1-matt@ozlabs.org/ v2: https://lore.kernel.org/all/20260527102319.100128-1-mattev@meta.com/ v1: https://lore.kernel.org/kvm/20260416131815.2729131-1-mattev@meta.com/ RFCv2: https://lore.kernel.org/kvm/20260312184613.3710705-1-mattev@meta.com/ RFCv1: https://lore.kernel.org/all/20260226202211.929005-1-mattev@meta.com/ Tech topic: https://lore.kernel.org/linux-iommu/20250918214425.2677057-1-amastro@fb.com/ Matt Evans (9): vfio/pci: Remove DMABUF export dependency on vdev->memory_lock vfio/pci: Un-revoke DMABUFs in LOW_POWER_ENTRY_WITH_WAKEUP resume dma-buf: Provide dma_buf_set_name() vfio/pci: Add a helper to look up PFNs for DMABUFs vfio/pci: Add a helper to create a DMABUF for a BAR-map VMA vfio/pci: Convert BAR mmap() to use a DMABUF vfio/pci: Clean up BAR zap and revocation vfio/pci: Support mmap() of a VFIO DMABUF vfio/pci: Revoke a DMABUF on request from userspace drivers/dma-buf/dma-buf.c | 84 ++- drivers/vfio/pci/Kconfig | 4 +- drivers/vfio/pci/Makefile | 3 +- .../vfio/pci/hisilicon/hisi_acc_vfio_pci.c | 14 +- drivers/vfio/pci/vfio_pci_config.c | 30 +- drivers/vfio/pci/vfio_pci_core.c | 222 +++++-- drivers/vfio/pci/vfio_pci_dmabuf.c | 609 +++++++++++++++--- drivers/vfio/pci/vfio_pci_priv.h | 54 +- include/linux/dma-buf.h | 16 +- include/linux/vfio_pci_core.h | 3 + include/uapi/linux/vfio.h | 24 + 11 files changed, 870 insertions(+), 193 deletions(-) -- 2.47.3

9 hours, 31 minutes

Re: [PATCH 3/3] dma-buf: heaps: cma: respect the device's maximum segment size

by Jason Gunthorpe

On Mon, Oct 05, 2026 at 08:46:40AM +0200, Karl Mehltretter wrote: > diff --git a/drivers/dma-buf/heaps/cma_heap.c b/drivers/dma-buf/heaps/cma_heap.c > index 3fb4b946c91a..9d61d14688b8 100644 > --- a/drivers/dma-buf/heaps/cma_heap.c > +++ b/drivers/dma-buf/heaps/cma_heap.c > @@ -58,16 +58,22 @@ static int cma_heap_attach(struct dma_buf *dmabuf, > { > struct cma_heap_buffer *buffer = dmabuf->priv; > struct dma_heap_attachment *a; > + unsigned int max_segment; > int ret; > > + max_segment = dma_get_max_seg_size(attachment->dev); > + /* The SG allocator requires a segment limit of at least PAGE_SIZE. */ > + if (max_segment < PAGE_SIZE) > + return -EINVAL; This seems like unnecessary AI paranoia.. If something like this should exist it should be inside sg_alloc_table_from_pages_segment(). I guess pretty much every DRM user has this latent issue? Given few importers even seem to use the segment size, it would be nicer to solve it by just not using segment size here. This is possible with the new dma api, and adding a new dma_buf_pages_to_sgt() using that approach looked OK from an AI draft. However, the dma_sync_sgtable_for_device() in the heap doesn't seem to have a straightforward conversion.. We never made streaming functions for the new dma api, and drivers are not permitted to call the cache flushing functions by themselves. Something to think about at least. Jason

2 days, 21 hours

Re: [RFC PATCH 1/2] iommu: Add iommu_map_sgtable_dma()

by Jason Gunthorpe

On Fri, Oct 09, 2026 at 02:21:13AM +0200, Karl Mehltretter wrote: > DMA-BUF attachments provide DMA addresses, not CPU-side pages. > iommu_map_sgtable() needs those pages, so private-domain importers such > as Rockchip cannot use it with strict DMABUF_DEBUG. > > Add an iommu-dma helper to map a live attachment into another domain by > translating through the device's default DMA domain. Roll back target > mappings on error. This still recovers physical addresses > internally. Absolutely not. You cannot recover the phys_addr's from a DMABUF expoter by using iommu_iova_to_phys(). Christian is adamant that drivers must not use physical addresses, and I will not agree to any hacks like this to try to end-run around his prohibition by abusing iommu subsystem. NAKed-by: Jason Gunthorpe <jgg(a)nvidia.com> If you want to keep the domain in the rockchip driver you have to fix DMABUF to allow an importer to use an iommu domain safely. It *must not* be done with a scatterlist interface, sorry, I know that is alot of work but Alistair is going to take a serious stab at it. Many users want this, not just rockchip. So help with the work to allow iommufd to do this and you will get it for free. Jason

3 days, 1 hour

Re: [RFC PATCH 2/3] dma-buf: add a warn-only mode to DMABUF_DEBUG

by Jason Gunthorpe

On Wed, Oct 07, 2026 at 02:02:31PM +0200, Diederik de Haas wrote: > [ 1091.981354] rc rc3: two consecutive events of type space > [ 1104.670513] input: EDIFIER e235 (AVRCP) as /devices/virtual/input/input12 > [ 1172.090231] devfreq fb000000.gpu: Couldn't update frequency transition information. > [ 1189.403278] devfreq fb000000.gpu: Couldn't update frequency transition information. > [ 1206.309295] input: EDIFIER e235 (AVRCP) as /devices/virtual/input/input13 > [ 1268.968696] DMA-BUF: importer used the CPU side of an exporter's sg_table This is a nice stack trace, is that the point of this series? > [ 1268.968710] CPU: 6 UID: 1000 PID: 29025 Comm: sway Not tainted 7.3-rc6+unreleased-arm64-cknow #1 PREEMPTLAZY Debian 7.3~rc6-3 > [ 1268.968715] Hardware name: FriendlyElec NanoPC-T6 Plus (DT) > [ 1268.968717] Call trace: > [ 1268.968719] show_stack+0x20/0x38 (C) > [ 1268.968726] dump_stack_lvl+0x60/0x80 > [ 1268.968730] sg_dmabuf_cpu_access_warn.part.0+0x24/0x30 > [ 1268.968734] sg_dmabuf_cpu_access_warn+0x34/0x38 > [ 1268.968739] iommu_map_sg+0xc8/0x1e0 > [ 1268.968745] rockchip_gem_iommu_map+0x8c/0x128 [rockchipdrm] > [ 1268.968757] rockchip_gem_prime_import_sg_table+0x58/0x160 [rockchipdrm] > [ 1268.968761] drm_gem_prime_import_dev+0xa8/0x1d0 [drm] > [ 1268.968776] drm_gem_prime_fd_to_handle+0x1a4/0x280 [drm] So.. This is the exact same thing I need for iommufd. rockchip is managing its own iommu domain and you cannot map to an iommu domain without using a physical address. Of course it is *completely* illegal to call iommu_map_sgtable() in the importer side of a dmabuf. static int rockchip_gem_iommu_map(struct rockchip_gem_object *rk_obj) { [..] ret = iommu_map_sgtable(private->domain, rk_obj->dma_addr, rk_obj->sgt, prot); Jason

3 days, 1 hour

Re: [RFC PATCH 0/2] iommu, drm/rockchip: Map private-domain imports from DMA addresses

by Robin Murphy

On 09/10/2026 1:21 am, Karl Mehltretter wrote: > Diederik reported failed video-buffer imports on RK3588 with > DMABUF_DEBUG. The proposed warn-only mode [1] confirmed CPU-side > attachment access in rockchip_gem_iommu_map() during Sway video resizing. > > This RFC keeps Rockchip's private scanout domain. Patch 1 copies live > iommu-dma mappings into another domain. Patch 2 uses it for PRIME > imports. Unsupported inputs keep the old page-based path, which remains > unfixed under strict DMABUF_DEBUG. > > Rob's MSM approach assumes direct DMA [2]. Rockchip's attachment instead > provides IOVAs from the VOP's default domain, which cannot be mapped > unchanged into the private domain. Each import retains both mappings > and adds one reverse lookup per 4 KiB. > > Christian rejected reverse translation in the earlier MSM proposal [3]. > Robin NAKed exporting iommu_get_dma_domain() [4]. Keeping the helper in > dma-iommu.c does not resolve the physical-address objection. And if I'd had the context of the whole series, I would have said what I can see now, that the entire premise is fundamentally wrong, irrespective of abusing the internal helper or not. > Sharing the first VOP's default DMA domain, as Exynos does, would avoid > the translation but also change native-buffer mapping. > > Is retaining the private domain and reverse-translating the DMA mappings > an acceptable direction? If so, I'll address the remaining limitations > before posting a non-RFC version. No. If a driver has attached the device to its own unmanaged IOMMU domain then it is using that domain, not the default domain, and thus has even less reason to go poking at the default domain than usual (where the "usual" is tenuous enough in itself). In this situation DMA mapping only needs to take care of non-coherent cache maintenance, and 32-bit ARM is actually the better example here. The fact that iommu-dma does a load of unnecessary work and returns a bogus DMA address just to still get the cache maintenance as a side-effect is a hideous inefficiency (which folks have complained about before...) and absolutely should not be relied upon. It needs to go away. There were reasons why in the original iommu_dma_ops design it was rather impractical to do better (in fact iommu_get_dma_domain() itself is largely just a hack around some of those limitations), but since b67483b3c44e ("iommu/dma: Centralise iommu_setup_dma_ops()") and particularly b5c58b2fdc42 ("dma-mapping: direct calls for dma-iommu"), it now really could and should be cleaned up - it just needs something slightly different from the standard dma-direct behaviour, as for this case we need to ignore the DMA mask and any bouncing conditions. Thanks, Robin. > Base: mainline 602042bf29f6. The warn-only RFC is not a prerequisite. > No stable backport requested. Strict-by-default DMABUF_DEBUG took effect > in v7.3-rc4. > > Testing (builds and QEMU only): > > - W=1 object and stub builds passed on arm64, ARM32 with/without LPAE, > x86-64 GCC/Clang, i386, s390, RISC-V and UML, including dynamic SWIOTLB. > - Rockchip strict/warn A/B reproduced the control failure and warning. > Treatments checked every page and byte in 83 imports each. Primary > buffers had 1/127/507 DMA segments under the default 64 KiB limit. > - SMMUv3 strict/lazy tests, missing-source, bounds and rollback tests > passed. Real bounced attachments exercised alignment and pool > rejection with no target mapping or DMA. > - Invalid forced-SWIOTLB Rockchip provider runs are excluded. Expected > segment-limit and unsupported-fallback diagnostics remain. > > The rig uses a custom Rockchip IOMMU model and the real GEM callback in a > test module, not VOP2, Sway or a real exporter. RK3588 hardware, two-VOP > and 32-bit ARM runtime testing are outstanding. > > Hardware testing is welcome, particularly with Diederik's Sway resize > workload. Compare control and both patches on the same base/config, > first with strict DMABUF_DEBUG, then with the warn-only RFC on both. > Please report full dmesg, config, exporter and visible display problems. > Keep a known-good boot kernel. DMABUF_DEBUG=n testing is welcome too. > > Developed and tested with LLM assistance. > > [1] https://lore.kernel.org/r/20261005064133.7305-1-kmehltretter@gmail.com/ > [2] https://lore.kernel.org/r/20261006131000.81501-1-robin.clark@oss.qualcomm.c… > [3] https://lore.kernel.org/r/bd4e5ece-1358-4e0b-bb04-ba9de62d26f6@amd.com/ > [4] https://lore.kernel.org/r/47bf9a4c-2a47-4e93-bcdf-8d953c9a5ab8@arm.com/ > > Reports: > https://lore.kernel.org/r/DLR74W1U9YPC.375IK0HOYHDIG@cknow-tech.com/ > https://lore.kernel.org/r/DLYLA3WQNN3X.3FB9MO8ZWBQ5@cknow-tech.com/ > > Karl Mehltretter (2): > iommu: Add iommu_map_sgtable_dma() > drm/rockchip: Map imported buffers from DMA addresses > > drivers/gpu/drm/rockchip/rockchip_drm_gem.c | 19 ++- > drivers/iommu/dma-iommu.c | 131 ++++++++++++++++++++ > include/linux/iommu.h | 12 ++ > 3 files changed, 157 insertions(+), 5 deletions(-) > > > base-commit: 602042bf29f6efde39cfb5fdd9289bf4854bc0c5

3 days, 2 hours

Jump to page:

2026

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

Linaro-mm-sig