* [PATCH v14 0/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent support
@ 2026-10-20 8:43 Oleksii Moisieiev
2026-10-20 8:43 ` [PATCH v14 1/6] xen/domctl: chain SCI handling before IOMMU in assign_device domctl Oleksii Moisieiev
` (5 more replies)
0 siblings, 6 replies; 10+ messages in thread
From: Oleksii Moisieiev @ 2026-10-20 8:43 UTC (permalink / raw)
To: xen-devel@lists.xenproject.org
Cc: Andrew Cooper, Anthony PERARD, Bertrand Marquis, Jan Beulich,
Juergen Gross, Julien Grall, Michal Orzel, Oleksii Moisieiev,
Roger Pau Monné, Stefano Stabellini, Volodymyr Babchuk
Inroducing patch series which includes implementation of the SCI SCMI
SMC multi-agent support.
This patch series follows RFC v5 [3] series which was introducing both
SCMI single-agent and multi-agent support. After the discussion it was
decided to split features and upstream singe-agent support first. This
feature is merged for now to v4.21-rc2.
I'm starting this patch series from v6 to save the discussion history
and don't break changes log.
Patch - xen/domctl: extend XEN_DOMCTL_assign_device to handle not
only iommu
- add chainged handling of assigned DT devices to support
access-controller functionality through SCI framework.
Change was done in two parts:
- call to sci_do_domctl() to do_domctl()
- update iommu_do_dt_domctl() to check for dt_device_is_protected()
and not fail if DT device is not protected by IOMMU
Patch - xen/arm: scmi: introduce SCI SCMI SMC multi-agent driver
- add Xen-specific SCMI container compatible `xen,sci`
under `/chosen/xen`; Xen binds only to the `arm,scmi-smc` inside it and
ignores other SCMI nodes (e.g. under `/firmware`).
- describe secondary agents as `xen,scmi-agent` children of an `agents`
node, with `arm,smc-id`, `shmem` and optional `reg` for the agent ID.
Discover the ID through BASE_DISCOVER_AGENT when `reg` is absent.
- keep shared-memory nodes under `xen_scmi_config`; platform RAM
reservations are described separately in `/reserved-memory`.
- each guest using SCMI supplies its agent_id (dom0 via
`xen,dom0-sci-agent-id` property in xen,sci compatible node,
toolstack via `arm_sci = "type=scmi_smc_multiagent,agent_id=..."`, dom0less
via `xen,sci_type` + `xen,sci-agent-id` in `xen,domain`).
- factor out SCMI generic definitions and shmem code.
- passthrough configuration for SCMI guests mirrors other HW passthrough.
Patch - docs: arm: add SCI SCMI SMC multi-agent driver docs
- document the Xen SCMI container under `/chosen/xen/xen_scmi_config` and the
mediator’s binding rules; update examples accordingly.
All Xen-specific SCMI configuration now lives under `/chosen/`
to keep host DT changes isolated while leaving the host `/firmware/scmi`
untouched for Dom0 consumption.
Code can be found at:
https://github.com/oleksiimoisieiev/xen/tree/scmi_ma_upstrv6
[1] RFC v2:
http://patchwork.kernel.org/project/xen-devel/cover/cover.1644341635.git.oleksii_moisieiev@epam.com/
[2] RFC v3:
https://patchwork.kernel.org/project/xen-devel/patch/20250311111618.1850927-1-grygorii_strashko@epam.com
[3] RFC v5:
https://lore.kernel.org/xen-devel/cover.1753184487.git.oleksii_moisieiev@epam.com/
[4] SCMI single-agent:
https://lore.kernel.org/xen-devel/cover.1756995595.git.oleksii_moisieiev@epam.com/
SCMI spec:
https://developer.arm.com/documentation/den0056/e/?lang=en
SCMI bindings:
https://web.git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/firmware/arm,scmi.yaml
https://web.git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/access-controllers/access-controllers.yaml
Reference EL3 FW:
RPI5: https://github.com/xen-troops/arm-trusted-firmware/commits/rpi5_dev/
Renesas v4h:
https://github.com/GrygiriiS/arm-trusted-firmware/commits/rcar_gen4_v2.7_v4x-scmi_upd/
base-commit: dbe60f244c (Update Xen to 4.21, 2025-02-21)
Changes in v14:
- replace the flat secondary-agent list and its cell-count property with
xen,scmi-agent child nodes under agents, using reg for an optional ID
- scope secondary-agent parsing to the management channel's xen,sci container
- resolve discovered IDs before insertion and reject duplicate or invalid IDs
- validate shared-memory phandles, addresses and nonzero page-aligned sizes
- retain shared-memory nodes under xen_scmi_config and document platform
reservations separately, following the review discussion
- update binding examples and commit description for the child-node format
- match the const-qualified SCI domain_sanitise_config callback interface
- describe secondary agents as xen,scmi-agent nodes under agents
- update host, discovery and dom0less examples for the child-node format
- clarify shared-memory placement and platform RAM reservations
Changes in v13:
- memcpy_{from,to}io(): align only the I/O pointer and access the
regular memory side through get_unaligned_le32()/put_unaligned_le32().
- document the contract of the helpers in xen/io.h
- reword the Arm implementation notes to state the concrete access
pattern
- drop the incorrect claim about avoiding endianness conversion from the
commit message; note instead that the _le32 helpers pair with
readl()/writel() so the byte sequence is preserved on either endianness
- fix typo in commit description
Changes in v12:
- rebase to the latest staging
- add R-Bs from Stefano and Ack from Jan
- add missed Ack.
- Add missed R-B
- correct the Emacs footer of both files, which asked for 8 column
tab indentation while the code is in 4 column Xen style
- rebase to the latest staging
- return EINVAL if xen,sci_type=scmi_smc_multiagent was set but
CONFIG_SCMI_SMC_MA not set
- put arm_sci_agent_id after v8r_el1_msa and before pad to preserve
order. Reduced pad from uint16_t to uint8_t to fill the gap (was 2
bytes before adding arm_sci_agent_id and left only 1, so used uint8_t
to preserve structure size)
- fix scmi_handle_call to use arm_smccc_guest_smc helper function
- update send_smc_message function to use new format of
arm_smccc_1_1_smc
- guard the Dom0 SCMI agent id setup in create_dom0() and MA related
funcitons with CONFIG_SCMI_SMC_MA
- fix the xen_scmi_config example in booting.txt, which placed properties
after subnodes and was rejected by dtc with "Properties must precede
subnodes"
Changes in v11:
- Fix agent_id documentation: clarify it applies to SCMI SMC
multi-agent support only, not plain SCMI SMC (reviewer feedback)
- Remove "non-zero" from agent_id description to match accepted
range [0..254]
- Remove "UINT8_MAX (255) is treated as invalid" from user
documentation as unnecessary implementation detail
- Add LIBXL_HAVE_SCMI_SMC_MULTIAGENT feature macro in libxl.h to
advertise the new scmi_smc_multiagent type and agent_id field
- Add agent_id validation in
libxl__arch_domain_build_info_setdefault() to reject invalid values at
the libxl level, not only in xl
Changes in v10:
- remove unused sci_do_domctl stub from sci.h
- removed extra include in memcpy-{to/from}io.c files
- Fix tabs in MAINTAINERS file
- remove duplicate SPDX tag from scmi-shmem.c
- add cast to ARM_SMCCC_INVALID_PARAMETER to settle the sign since
ARM_SMCCC_INVALID_PARAMETER is -3 which is part of the spec and resp
is the default smccc call structure.
- update free_channel_list. Add spinlock to avoid race condition and
a comment with a description of the function work
- preserve error of smc_create_channel in scmi_probe
- check scmi shmem address alignment as wel as it is done for size
- check for d->arch.sci_data != NULL in scmi_handle_call
- use SCMI_SHMEM_MAPPED size for iomem_permit_access
- change len type to unsigned in shmem_{get|put}_message
- rename shmem_channel_is_free to shmem_channel_status
- add comment about skipping message status when getting message response
- Set correct agent_id ranges for dom0less and a toolstack. Agent_id 0
is not binded for dom0 so can be reused. Also mentioned that
UINT8_MAX (255) is treated as invalid agent_id.
- Split hypervisor and toolstack changes into separate commits
- move init list and spin init after initial checks in probe call
- fix typo in comments
- clean resources when sci_register returns an error.
- Split hypervisor and toolstack changes into separate commits
- rephrase section about /firmware/scmi. Mentioned that this node is
taken from Host DT and copied unmodified.
- fix xen,reg address for secondary domains for Dom0less configuration
Changes in v9:
- treat SCI as a gate for XEN_DOMCTL_*assign_device: abort before
IOMMU if sci_do_domctl() returns an error other than -ENXIO, instead
of trying to propagate SCI errors after a successful IOMMU
operation. This avoids partial success and the need for IOMMU rollback.
- remove early return from do_domctl() in the assign_device
path to keep RCU handling intact.
- change IS_ENABLED(*) to #ifdef in sci_do_domctl quard
- reword commit description to refer to memcpy_fromio and memcpy_toio
- ordering obj-y in Makefile
- rename ALL_LIBS to ARCH_LIBS
- drop io.h and move definitions to the common header, fix comments to
be arch neutral
- update comments for memcpy_{from/to}io implementation
- sort and refactor MAINTAINERS enties
- remove Spurious changes
- add extra check to avoid ASSERT when calling unmap_channel_memory
from assign device method
- set correct tx flag to SCMI_BASE_AGENT_PERMISSIONS_RESET when
freeing resources. Flag should be set to 1 according to the
section 4.2.2.12 [0].
- fix dt node copmaring
- moved channel->shmem check from ASSERT in unmap_memory_channel to
"if" statement. This will prevent firing ASSERT if
unmap_channel_memory was called twice on the same channel.
Changes in v8:
- check for CONFIG_ARM_SCI to be ebabled instead of COMFIG_ARM before
calling sci_do_domctl
- rework sci_do_domctl call to avoid extra checks, improved error
handling.
- do not propagate ret1 if sci_do_domctl returned positive ret
- updated comment in domctl.c code
- switched to ordered accessors to address the ordering and barrier
concerns.
- updated the documentation to match the implementation and explicitly
state the supported access sizes and granularity.
- rename memcpy_* implementation files to memcpu-* to follow naming
convension
- fix indentation to match Xen style
- fix intendation to match Xen style
- move memcpy-{from/to}io to more convenient library place
- update xen_scmi func_id in commit description
- updated documentation with the new DT format
- updated opt_dom0_scmi_agent_id setting to avoid it to be equal
SCMI_AGENT_ID_INVALID.
- changed SCMI_AGENT_ID_INVALID from 0xff to UINT8_MAX which makes
code more clear showing that UINT8_MAX is theated like invalid
agent_id and couldn't be used. Also excluded SCMI_AGENT_ID_INVALID
from acceptable value range
- remove outdated xen,config property ignore, added xen,sci compatible
to skip_matches in handle_node
- add documentation for pre-existing scmi-smc-passthrough command line
option in alphabetically correct location (in 's' section)
- add note to commit description about documentation for previously
undocumented scmi-smc-passthrough
- Fix SMC IDs in DT examples (Xen management uses 0x82000003, Dom0 uses 0x82000002)
- Add explicit note explaining why Dom0 and Xen channels do not conflict
- Document dom0less multi-agent configuration example (xen,sci_type / xen,sci-agent-id)
- Add scmi_xen node to agent-discovery example with #scmi-secondary-agents-cells = 2
- Drop dom0=sci-agent-id command line handling; Dom0 SCMI is now enabled via
xen,dom0-sci-agent-id in the xen,sci DT container
- Refresh docs and examples to mention the DT property instead of the cmdline option
- update documentation to match the last DT format
- fixed RST: "... code-block:: dts" -> ".. code-block:: dts"
- update documentation with dom0less configuration example
- update documentation with new param xen,dom0-sci-agent-id
instead of the command line parameter
Changes in v7:
- update domctl to build on both Arm and x86 platforms
- move ret1 declaration to the top of the function as required by code
style
- x86 guidance: removed the speculative note; header now just says
each arch supplies its own implementation or macro.
- name spacing: dropped the double-underscore; the helpers are now
memcpy_fromio / memcpy_toio. The header also explicitly allows an
arch to define these as macros before including it.
- updated io.c to keep 32-bit transfers safe on arm32
- moved to __raw_read*/__raw_write* accessors to avoid endianness conversion.
- split the helpers into separate compilation units
- rework scmi nodes for xen to match on compatible string instead of
the direct path
- update documentation in section of the xen_scmi configuration which
is matched by "xen,sci" compatible instead of the direct path.
Changes in v6:
- change iommu_do_domctl and sci_do_domctl command order and
call sci_do_domctl first which will produce cleaner code path.
Also dropped changing return code when iommu was disabled in
iommu_do_domctl.
- sorted objs in Makefile alhabetically
- added newline at the end of Makefile
- used uint{N}_t intead of u{N}
- add comment about why 32 bit IO operations were used
- updated cast opertaions to avoid dropping constness which is wrong
- move function definitions to generic place so the could be reused by
other arch
- add SPDX tag to io.c
- updated scmi-shmem to use io.h from generic location
- update scmi_agent_id parameter to be provided inside dom0= parameter
list and have the following format "dom0=sci-agent-id=0"
This change was done as a response for Stefano comment and
requires a lot of code changes, but produces much cleaner solution
that's why I've added it to the code.
- fix file comments and return codes
- fix lenght checks in shmem_{get,put}_message to use offsetof
- remove len member from scmi_channel structure as it is not used
- set scmi-secondary-agents property to be mandatory since if no
secondary agents were provided then there is no sence to enable scmi
when no secondary agents are populated to the Domains
- update documentation in booting.txt, added xen_scmi node to the
example
- adjust d->arch.sci_enabled value in scmi_domain_destroy
- fix lock management in smc_create_channel call
- avoid extra map_channel_memory command for Xen management channel
because collect_agent_id call unmaps memory if DOMID_XEN is not
set. So for Xen management channel we can init domain_id ad DOMID_XEN
before calling collect_agent_id so memory shouldn't be unmapped.
- remove all HVC mentions from the multi-agent doc
- update sci-agent-id parameter description in the documentation
- add missing Sign-of
- minor fixes across the document
Changes in v5:
- return -EINVAL if mediator without assign_dt_device was provided
- invert return code check for iommu_do_domctl in
XEN_DOMCTL_assign_device domctl processing to make cleaner code
- change -ENOTSUPP error code to -ENXIO in sci_do_domctl
- handle -ENXIO return comde of iommu_do_domctl
- leave !dt_device_is_protected check in iommu_do_dt_domctl to make
code work the same way it's done in "handle_device" call while
creating hwdom(dom0) and "handle_passthrough_prop" call for dom0less
creation
- drop return check from sci_assign_dt_device call as not needed
- do not return EINVAL when addign_dt_device is not set. That is
because this callback is optional and not implemented in single-agent driver
- move memcpy_toio/fromio to the generic place
- fix device-tree example format in booting.txt, added ";" after "}".
- update define in scmi-proto.h
- update define in scmi-shmem.h file
- scmi_assign_device - do not ignore -EOPNOTSUPP return
code of the do_smc_xfer
- remove overwriting agent_channel->agent_id after
SCMI_BASE_DISCOVER_AGENT call
- add multi-agent files to the MAINTAINERS
- add SCMI multi-agent description to the SUPPORT.md
- handle ARM_SMCCC_INVALID_PARAMETER return code and return -EINVAL
for smc call
- updated collect_agents function. Set agent_id parameter as optional
in scmi-secondary-agents device-tree property
- introduce "#scmi-secondary-agents-cells" parameter to set if
agent_id was provided
- reanme xen,scmi-secondary-agents property to scmi-secondary-agents
- move memcpu_toio/fromio for the generic place
- update Xen to get management channel from /chosen/xen,config node
- get hypervisor channnel from node instead of using hardcoded
- update handling scmi and shmem nodes for the domain
- Set multi-agent driver to support only Arm64
- rework multi-agent driver to leave Host Device-tree unmodified
Changes in v4:
- toolstack comments from Anthony PERARD
- added dom0less support
- added doc for "xen,scmi-secondary-agents"
Grygorii Strashko (2):
xen/domctl: chain SCI handling before IOMMU in assign_device domctl
docs: arm: add SCI SCMI SMC multi-agent driver docs
Oleksii Moisieiev (4):
xen: arm: smccc: add INVALID_PARAMETER error code
lib/arm: Add I/O memory copy helpers
xen/arm: scmi: introduce SCI SCMI SMC multi-agent driver
tools/xl/libxl: wire up SCMI SMC multi-agent configuration
MAINTAINERS | 1 +
SUPPORT.md | 11 +
.../arm/firmware/arm-scmi.rst | 405 +++++++++
docs/man/xl.cfg.5.pod.in | 13 +
docs/misc/arm/device-tree/booting.txt | 195 +++++
tools/include/libxl.h | 8 +
tools/libs/light/libxl_arm.c | 13 +
tools/libs/light/libxl_types.idl | 4 +-
tools/xl/xl_parse.c | 27 +
xen/arch/arm/Makefile | 1 +
xen/arch/arm/arch.mk | 1 +
xen/arch/arm/dom0less-build.c | 17 +
xen/arch/arm/domain_build.c | 43 +
xen/arch/arm/firmware/Kconfig | 12 +
xen/arch/arm/firmware/Makefile | 1 +
xen/arch/arm/firmware/sci.c | 36 +
xen/arch/arm/firmware/scmi-proto.h | 164 ++++
xen/arch/arm/firmware/scmi-shmem.c | 118 +++
xen/arch/arm/firmware/scmi-shmem.h | 45 +
xen/arch/arm/firmware/scmi-smc-multiagent.c | 792 ++++++++++++++++++
xen/arch/arm/include/asm/firmware/sci.h | 8 +
xen/arch/arm/include/asm/smccc.h | 1 +
xen/arch/arm/lib/Makefile | 2 +
xen/arch/arm/lib/memcpy-fromio.c | 56 ++
xen/arch/arm/lib/memcpy-toio.c | 56 ++
xen/common/domctl.c | 15 +
xen/drivers/passthrough/device_tree.c | 6 +
xen/include/public/arch-arm.h | 5 +-
xen/include/xen/io.h | 25 +
29 files changed, 2079 insertions(+), 2 deletions(-)
create mode 100644 xen/arch/arm/firmware/scmi-proto.h
create mode 100644 xen/arch/arm/firmware/scmi-shmem.c
create mode 100644 xen/arch/arm/firmware/scmi-shmem.h
create mode 100644 xen/arch/arm/firmware/scmi-smc-multiagent.c
create mode 100644 xen/arch/arm/lib/Makefile
create mode 100644 xen/arch/arm/lib/memcpy-fromio.c
create mode 100644 xen/arch/arm/lib/memcpy-toio.c
--
2.43.0
base-commit: aa73a1b99309f22d9bbc2603d1b206e7298d9473
branch: scmi_ma_upstrv14
^ permalink raw reply [flat|nested] 10+ messages in thread* [PATCH v14 1/6] xen/domctl: chain SCI handling before IOMMU in assign_device domctl 2026-10-20 8:43 [PATCH v14 0/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent support Oleksii Moisieiev @ 2026-10-20 8:43 ` Oleksii Moisieiev 2026-10-20 8:43 ` [PATCH v14 2/6] xen: arm: smccc: add INVALID_PARAMETER error code Oleksii Moisieiev ` (4 subsequent siblings) 5 siblings, 0 replies; 10+ messages in thread From: Oleksii Moisieiev @ 2026-10-20 8:43 UTC (permalink / raw) To: xen-devel@lists.xenproject.org Cc: Andrew Cooper, Anthony PERARD, Bertrand Marquis, Jan Beulich, Juergen Gross, Julien Grall, Michal Orzel, Oleksii Moisieiev, Roger Pau Monné, Stefano Stabellini, Volodymyr Babchuk, Grygorii Strashko From: Grygorii Strashko <grygorii_strashko@epam.com> Add chained handling of assigned DT devices to support access-controller functionality through SCI framework, so a DT device assign request can be passed to firmware for processing and enabling VM access to the requested device (for example, device power management through SCMI). The SCI access-controller DT device processing is called before the IOMMU path. It runs for any DT-described device (protected or not, and even when the IOMMU is disabled). The IOMMU path remains unchanged for PCI devices; only the DT path is relaxed to permit non-IOMMU devices. This lets xl.cfg:"dtdev" list both IOMMU-protected and non-protected DT devices: dtdev = [ "/soc/video@e6ef0000", <- IOMMU protected device "/soc/i2c@e6508000", <- not IOMMU protected device ] The change is done in two parts: 1) call sci_do_domctl() in do_domctl() before IOMMU processing. If sci_do_domctl() reports an error other than -ENXIO, treat it as authoritative and skip the IOMMU path. A return of -ENXIO indicates that SCI did not handle the request and is ignored, allowing the existing IOMMU handling to run unchanged; 2) update iommu_do_dt_domctl() to check for dt_device_is_protected() and not fail if DT device is not protected by IOMMU. iommu_do_pci_domctl doesn't need to be updated because iommu_do_domctl first tries iommu_do_pci_domctl (when CONFIG_HAS_PCI) and falls back to iommu_do_dt_domctl only if PCI returns -ENODEV. The new dt_device_is_protected() bypass in iommu_do_dt_domctl only applies to DT-described devices; SCI parameters are carried via DT nodes. PCI devices handled by iommu_do_pci_domctl do not carry DT/SCI metadata in this path, so there is no notion of “SCI parameters on a non-IOMMU-protected PCI device” for it to interpret or to skip. The PCI path should continue to report errors if assignment cannot be performed by the IOMMU layer. So we should leave iommu_do_pci_domctl unchanged; the SCI/DT-specific relaxations belong only in the DT path. Also SCI handling only exists when DT is present. Signed-off-by: Grygorii Strashko <grygorii_strashko@epam.com> Signed-off-by: Oleksii Moisieiev <oleksii_moisieiev@epam.com> Reviewed-by: Stefano Stabellini <sstabellini@kernel.org> Acked-by: Jan Beulich <jbeulich@suse.com> --- (no changes since v12) Changes in v12: - rebase to the latest staging - add R-Bs from Stefano and Ack from Jan Changes in v10: - remove unused sci_do_domctl stub from sci.h Changes in v9: - treat SCI as a gate for XEN_DOMCTL_*assign_device: abort before IOMMU if sci_do_domctl() returns an error other than -ENXIO, instead of trying to propagate SCI errors after a successful IOMMU operation. This avoids partial success and the need for IOMMU rollback. - remove early return from do_domctl() in the assign_device path to keep RCU handling intact. - change IS_ENABLED(*) to #ifdef in sci_do_domctl quard Changes in v8: - check for CONFIG_ARM_SCI to be ebabled instead of COMFIG_ARM before calling sci_do_domctl - rework sci_do_domctl call to avoid extra checks, improved error handling. - do not propagate ret1 if sci_do_domctl returned positive ret - updated comment in domctl.c code Changes in v7: - update domctl to build on both Arm and x86 platforms - move ret1 declaration to the top of the function as required by code style Changes in v6: - change iommu_do_domctl and sci_do_domctl command order and call sci_do_domctl first which will produce cleaner code path. Also dropped changing return code when iommu was disabled in iommu_do_domctl. Changes in v5: - return -EINVAL if mediator without assign_dt_device was provided - invert return code check for iommu_do_domctl in XEN_DOMCTL_assign_device domctl processing to make cleaner code - change -ENOTSUPP error code to -ENXIO in sci_do_domctl - handle -ENXIO return comde of iommu_do_domctl - leave !dt_device_is_protected check in iommu_do_dt_domctl to make code work the same way it's done in "handle_device" call while creating hwdom(dom0) and "handle_passthrough_prop" call for dom0less creation - drop return check from sci_assign_dt_device call as not needed - do not return EINVAL when addign_dt_device is not set. That is because this callback is optional and not implemented in single-agent driver xen/arch/arm/firmware/sci.c | 36 +++++++++++++++++++++++++ xen/arch/arm/include/asm/firmware/sci.h | 8 ++++++ xen/common/domctl.c | 15 +++++++++++ xen/drivers/passthrough/device_tree.c | 6 +++++ 4 files changed, 65 insertions(+) diff --git a/xen/arch/arm/firmware/sci.c b/xen/arch/arm/firmware/sci.c index f73ed06092..371eca019c 100644 --- a/xen/arch/arm/firmware/sci.c +++ b/xen/arch/arm/firmware/sci.c @@ -126,6 +126,42 @@ int sci_assign_dt_device(struct domain *d, struct dt_device_node *dev) return 0; } +int sci_do_domctl(struct xen_domctl *domctl, struct domain *d, + XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl) +{ + struct dt_device_node *dev; + int ret = 0; + + switch ( domctl->cmd ) + { + case XEN_DOMCTL_assign_device: + ret = -ENXIO; + if ( domctl->u.assign_device.dev != XEN_DOMCTL_DEV_DT ) + break; + + if ( !cur_mediator ) + break; + + if ( !cur_mediator->assign_dt_device ) + break; + + ret = dt_find_node_by_gpath(domctl->u.assign_device.u.dt.path, + domctl->u.assign_device.u.dt.size, &dev); + if ( ret ) + return ret; + + ret = sci_assign_dt_device(d, dev); + + break; + + default: + /* do not fail here as call is chained with iommu handling */ + break; + } + + return ret; +} + static int __init sci_init(void) { struct dt_device_node *np; diff --git a/xen/arch/arm/include/asm/firmware/sci.h b/xen/arch/arm/include/asm/firmware/sci.h index 1d566be8e2..6d8286b5cb 100644 --- a/xen/arch/arm/include/asm/firmware/sci.h +++ b/xen/arch/arm/include/asm/firmware/sci.h @@ -146,6 +146,14 @@ int sci_dt_finalize(struct domain *d, void *fdt); * control" functionality. */ int sci_assign_dt_device(struct domain *d, struct dt_device_node *dev); + +/* + * SCI domctl handler + * + * Only XEN_DOMCTL_assign_device is handled for now. + */ +int sci_do_domctl(struct xen_domctl *domctl, struct domain *d, + XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl); #else #include <public/arch-arm.h> diff --git a/xen/common/domctl.c b/xen/common/domctl.c index 83405a766a..1d5b92631c 100644 --- a/xen/common/domctl.c +++ b/xen/common/domctl.c @@ -29,6 +29,9 @@ #include <xen/xvmalloc.h> #include <asm/current.h> +#ifdef CONFIG_ARM +#include <asm/firmware/sci.h> +#endif #include <asm/irq.h> #include <asm/page.h> #include <asm/p2m.h> @@ -925,6 +928,18 @@ long do_domctl(XEN_GUEST_HANDLE_PARAM(xen_domctl_t) u_domctl) case XEN_DOMCTL_assign_device: case XEN_DOMCTL_test_assign_device: case XEN_DOMCTL_deassign_device: + /* + * Chain SCI DT handling ahead of the IOMMU path so an SCI mediator + * can authorise access-controlled DT devices. Unhandled cases report + * -ENXIO, which is ignored. Any other SCI error aborts before the + * IOMMU path runs. + */ +#ifdef CONFIG_ARM_SCI + ret = sci_do_domctl(op, d, u_domctl); + if ( ret < 0 && ret != -ENXIO ) + break; +#endif + ret = iommu_do_domctl(op, d, u_domctl); break; diff --git a/xen/drivers/passthrough/device_tree.c b/xen/drivers/passthrough/device_tree.c index ab1e07ac99..7c4cf27d32 100644 --- a/xen/drivers/passthrough/device_tree.c +++ b/xen/drivers/passthrough/device_tree.c @@ -379,6 +379,12 @@ int iommu_do_dt_domctl(struct xen_domctl *domctl, struct domain *d, break; } + if ( !dt_device_is_protected(dev) ) + { + ret = 0; + break; + } + ret = iommu_assign_dt_device(d, dev); if ( ret ) -- 2.43.0 ^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH v14 2/6] xen: arm: smccc: add INVALID_PARAMETER error code 2026-10-20 8:43 [PATCH v14 0/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent support Oleksii Moisieiev 2026-10-20 8:43 ` [PATCH v14 1/6] xen/domctl: chain SCI handling before IOMMU in assign_device domctl Oleksii Moisieiev @ 2026-10-20 8:43 ` Oleksii Moisieiev 2026-10-20 8:43 ` [PATCH v14 3/6] lib/arm: Add I/O memory copy helpers Oleksii Moisieiev ` (3 subsequent siblings) 5 siblings, 0 replies; 10+ messages in thread From: Oleksii Moisieiev @ 2026-10-20 8:43 UTC (permalink / raw) To: xen-devel@lists.xenproject.org Cc: Andrew Cooper, Anthony PERARD, Bertrand Marquis, Jan Beulich, Juergen Gross, Julien Grall, Michal Orzel, Oleksii Moisieiev, Roger Pau Monné, Stefano Stabellini, Volodymyr Babchuk According to the "7.1 Return Codes" section of DEN0028 [1] INVALID_PARAMETER code (-3) is returned when one of the call parameters has a non-supported value. Adding this error code to the common smccc header file. [1]: https://documentation-service.arm.com/static/5f8edaeff86e16515cdbe4c6 Signed-off-by: Oleksii Moisieiev <oleksii_moisieiev@epam.com> Acked-by: Stefano Stabellini <sstabellini@kernel.org> --- (no changes since v12) Changes in v12: - add missed Ack. xen/arch/arm/include/asm/smccc.h | 1 + 1 file changed, 1 insertion(+) diff --git a/xen/arch/arm/include/asm/smccc.h b/xen/arch/arm/include/asm/smccc.h index c7763acd7f..91c9bc1186 100644 --- a/xen/arch/arm/include/asm/smccc.h +++ b/xen/arch/arm/include/asm/smccc.h @@ -381,6 +381,7 @@ void arm_smccc_1_2_smc(const struct arm_smccc_1_2_regs *args, 0x3FFF) /* SMCCC error codes */ +#define ARM_SMCCC_INVALID_PARAMETER (-3) #define ARM_SMCCC_NOT_REQUIRED (-2) #define ARM_SMCCC_ERR_UNKNOWN_FUNCTION (-1) #define ARM_SMCCC_NOT_SUPPORTED (-1) -- 2.43.0 ^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH v14 3/6] lib/arm: Add I/O memory copy helpers 2026-10-20 8:43 [PATCH v14 0/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent support Oleksii Moisieiev 2026-10-20 8:43 ` [PATCH v14 1/6] xen/domctl: chain SCI handling before IOMMU in assign_device domctl Oleksii Moisieiev 2026-10-20 8:43 ` [PATCH v14 2/6] xen: arm: smccc: add INVALID_PARAMETER error code Oleksii Moisieiev @ 2026-10-20 8:43 ` Oleksii Moisieiev 2026-10-20 8:43 ` [PATCH v14 4/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent driver Oleksii Moisieiev ` (2 subsequent siblings) 5 siblings, 0 replies; 10+ messages in thread From: Oleksii Moisieiev @ 2026-10-20 8:43 UTC (permalink / raw) To: xen-devel@lists.xenproject.org Cc: Andrew Cooper, Anthony PERARD, Bertrand Marquis, Jan Beulich, Juergen Gross, Julien Grall, Michal Orzel, Oleksii Moisieiev, Roger Pau Monné, Stefano Stabellini, Volodymyr Babchuk Introduce memcpy_fromio() and memcpy_toio() helpers to copy between regular memory and a memory-like I/O region on Arm. The generic prototypes live in io.h so other architectures can provide their own implementations. The helpers copy a byte sequence: byte accesses are used until the I/O pointer is 32-bit aligned and for the trailing count % 4 bytes, the aligned bulk uses 32-bit accesses. Only the I/O pointer is aligned, the regular memory side is accessed through get_unaligned_le32() / put_unaligned_le32(), so the access widths issued on the I/O side do not depend on the alignment of the memory buffer. These _le32 helpers pair with readl()/writel(), which are little endian accessors, hence the byte sequence is preserved on either CPU endianness. The constraints on the I/O region are documented next to the declarations in xen/io.h, the arch-specific access pattern next to the implementation. Signed-off-by: Oleksii Moisieiev <oleksii_moisieiev@epam.com> Reviewed-by: Stefano Stabellini <sstabellini@kernel.org> --- (no changes since v13) Changes in v13: - memcpy_{from,to}io(): align only the I/O pointer and access the regular memory side through get_unaligned_le32()/put_unaligned_le32(). - document the contract of the helpers in xen/io.h - reword the Arm implementation notes to state the concrete access pattern - drop the incorrect claim about avoiding endianness conversion from the commit message; note instead that the _le32 helpers pair with readl()/writel() so the byte sequence is preserved on either endianness Changes in v12: - Add missed R-B - correct the Emacs footer of both files, which asked for 8 column tab indentation while the code is in 4 column Xen style Changes in v10: - removed extra include in memcpy-{to/from}io.c files Changes in v9: - reword commit description to refer to memcpy_fromio and memcpy_toio - ordering obj-y in Makefile - rename ALL_LIBS to ARCH_LIBS - drop io.h and move definitions to the common header, fix comments to be arch neutral - update comments for memcpy_{from/to}io implementation Changes in v8: - switched to ordered accessors to address the ordering and barrier concerns. - updated the documentation to match the implementation and explicitly state the supported access sizes and granularity. - rename memcpy_* implementation files to memcpu-* to follow naming convension - fix indentation to match Xen style - fix intendation to match Xen style - move memcpy-{from/to}io to more convenient library place Changes in v7: - x86 guidance: removed the speculative note; header now just says each arch supplies its own implementation or macro. - name spacing: dropped the double-underscore; the helpers are now memcpy_fromio / memcpy_toio. The header also explicitly allows an arch to define these as macros before including it. - updated io.c to keep 32-bit transfers safe on arm32 - moved to __raw_read*/__raw_write* accessors to avoid endianness conversion. - split the helpers into separate compilation units Changes in v6: - sorted objs in Makefile alhabetically - added newline at the end of Makefile - used uint{N}_t intead of u{N} - add comment about why 32 bit IO operations were used - updated cast opertaions to avoid dropping constness which is wrong - move function definitions to generic place so the could be reused by other arch - add SPDX tag to io.c Changes in v5: - move memcpy_toio/fromio to the generic place xen/arch/arm/Makefile | 1 + xen/arch/arm/arch.mk | 1 + xen/arch/arm/lib/Makefile | 2 ++ xen/arch/arm/lib/memcpy-fromio.c | 56 ++++++++++++++++++++++++++++++++ xen/arch/arm/lib/memcpy-toio.c | 56 ++++++++++++++++++++++++++++++++ xen/include/xen/io.h | 25 ++++++++++++++ 6 files changed, 141 insertions(+) create mode 100644 xen/arch/arm/lib/Makefile create mode 100644 xen/arch/arm/lib/memcpy-fromio.c create mode 100644 xen/arch/arm/lib/memcpy-toio.c diff --git a/xen/arch/arm/Makefile b/xen/arch/arm/Makefile index cdcec8cd03..ba61f9e5fd 100644 --- a/xen/arch/arm/Makefile +++ b/xen/arch/arm/Makefile @@ -8,6 +8,7 @@ ifneq ($(CONFIG_NO_PLAT),y) obj-y += platforms/ endif obj-y += firmware/ +obj-y += lib/ obj-$(CONFIG_TEE) += tee/ obj-$(CONFIG_HAS_VPCI) += vpci.o diff --git a/xen/arch/arm/arch.mk b/xen/arch/arm/arch.mk index dea8dbd18a..009bb22c45 100644 --- a/xen/arch/arm/arch.mk +++ b/xen/arch/arm/arch.mk @@ -2,6 +2,7 @@ # arm-specific definitions ARCH_LIBS-y += arch/arm/$(ARCH)/lib/lib.a +ARCH_LIBS-y += arch/arm/lib/lib.a $(call cc-options-add,CFLAGS,CC,$(EMBEDDED_EXTRA_CFLAGS)) $(call cc-option-add,CFLAGS,CC,-Wnested-externs) diff --git a/xen/arch/arm/lib/Makefile b/xen/arch/arm/lib/Makefile new file mode 100644 index 0000000000..07a0d9186c --- /dev/null +++ b/xen/arch/arm/lib/Makefile @@ -0,0 +1,2 @@ +lib-y += memcpy-fromio.o +lib-y += memcpy-toio.o diff --git a/xen/arch/arm/lib/memcpy-fromio.c b/xen/arch/arm/lib/memcpy-fromio.c new file mode 100644 index 0000000000..98339a5cfa --- /dev/null +++ b/xen/arch/arm/lib/memcpy-fromio.c @@ -0,0 +1,56 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ + +#include <xen/io.h> +#include <xen/unaligned.h> + +/* + * Arm implementation notes: + * - Byte accesses are issued until the I/O pointer is 32-bit aligned and + * for the trailing count % 4 bytes; the aligned bulk uses 32-bit + * accesses. No wider accesses are issued. + * - The access widths on the I/O side depend only on the alignment of the + * I/O pointer and on count. The regular memory side is accessed through + * put_unaligned_le32(), so its alignment has no influence on them. + * - put_unaligned_le32() pairs with readl(), which is a little endian + * accessor, so the byte sequence is preserved on either CPU endianness. + * - The I/O region must be mapped with device attributes; ordering is + * provided by the accessors themselves, no extra barriers are added. + */ + +void memcpy_fromio(void *to, const volatile void __iomem *from, + size_t count) +{ + while ( count && !IS_ALIGNED((unsigned long)from, 4) ) + { + *(uint8_t *)to = readb(from); + from++; + to++; + count--; + } + + while ( count >= 4 ) + { + put_unaligned_le32(readl(from), to); + from += 4; + to += 4; + count -= 4; + } + + while ( count ) + { + *(uint8_t *)to = readb(from); + from++; + to++; + count--; + } +} + +/* + * Local variables: + * mode: C + * c-file-style: "BSD" + * c-basic-offset: 4 + * tab-width: 4 + * indent-tabs-mode: nil + * End: + */ diff --git a/xen/arch/arm/lib/memcpy-toio.c b/xen/arch/arm/lib/memcpy-toio.c new file mode 100644 index 0000000000..c0be79586d --- /dev/null +++ b/xen/arch/arm/lib/memcpy-toio.c @@ -0,0 +1,56 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ + +#include <xen/io.h> +#include <xen/unaligned.h> + +/* + * Arm implementation notes: + * - Byte accesses are issued until the I/O pointer is 32-bit aligned and + * for the trailing count % 4 bytes; the aligned bulk uses 32-bit + * accesses. No wider accesses are issued. + * - The access widths on the I/O side depend only on the alignment of the + * I/O pointer and on count. The regular memory side is accessed through + * get_unaligned_le32(), so its alignment has no influence on them. + * - get_unaligned_le32() pairs with writel(), which is a little endian + * accessor, so the byte sequence is preserved on either CPU endianness. + * - The I/O region must be mapped with device attributes; ordering is + * provided by the accessors themselves, no extra barriers are added. + */ + +void memcpy_toio(volatile void __iomem *to, const void *from, + size_t count) +{ + while ( count && !IS_ALIGNED((unsigned long)to, 4) ) + { + writeb(*(const uint8_t *)from, to); + from++; + to++; + count--; + } + + while ( count >= 4 ) + { + writel(get_unaligned_le32(from), to); + from += 4; + to += 4; + count -= 4; + } + + while ( count ) + { + writeb(*(const uint8_t *)from, to); + from++; + to++; + count--; + } +} + +/* + * Local variables: + * mode: C + * c-file-style: "BSD" + * c-basic-offset: 4 + * tab-width: 4 + * indent-tabs-mode: nil + * End: + */ diff --git a/xen/include/xen/io.h b/xen/include/xen/io.h index 164a20c5d7..0b60be27c1 100644 --- a/xen/include/xen/io.h +++ b/xen/include/xen/io.h @@ -67,4 +67,29 @@ static inline bool write_mmio(volatile void __iomem *mem, unsigned long data, return true; } +/* + * Copy a sequence of bytes between regular memory and a memory-like I/O + * region, e.g. shared memory placed in RAM or SRAM mapped with device + * attributes. + * + * The I/O region must behave like byte-addressable storage: it has to + * accept 8-bit accesses at any byte address and naturally aligned 32-bit + * accesses, with plain byte-storage semantics in both cases (no side + * effects, no dependency on the access width). An implementation may use + * any of these widths, but never wider ones. The helpers are therefore + * not suitable for device registers with access-width requirements. + * + * Neither pointer needs to be aligned. The access widths issued on the + * I/O side depend only on the alignment of the I/O pointer and on count, + * never on the alignment of the regular memory pointer. + * + * Implementations are architecture-specific and may impose further + * constraints; see the respective arch/<arch>/lib/memcpy-{from,to}io.c + * for the exact access pattern. + */ +void memcpy_fromio(void *to, const volatile void __iomem *from, + size_t count); +void memcpy_toio(volatile void __iomem *to, const void *from, + size_t count); + #endif /* XEN_IO_H */ -- 2.43.0 ^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH v14 4/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent driver 2026-10-20 8:43 [PATCH v14 0/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent support Oleksii Moisieiev ` (2 preceding siblings ...) 2026-10-20 8:43 ` [PATCH v14 3/6] lib/arm: Add I/O memory copy helpers Oleksii Moisieiev @ 2026-10-20 8:43 ` Oleksii Moisieiev 2026-10-23 1:56 ` Volodymyr Babchuk 2026-10-20 8:43 ` [PATCH v14 6/6] docs: arm: add SCI SCMI SMC multi-agent driver docs Oleksii Moisieiev 2026-10-20 8:43 ` [PATCH v14 5/6] tools/xl/libxl: wire up SCMI SMC multi-agent configuration Oleksii Moisieiev 5 siblings, 1 reply; 10+ messages in thread From: Oleksii Moisieiev @ 2026-10-20 8:43 UTC (permalink / raw) To: xen-devel@lists.xenproject.org Cc: Andrew Cooper, Anthony PERARD, Bertrand Marquis, Jan Beulich, Juergen Gross, Julien Grall, Michal Orzel, Oleksii Moisieiev, Roger Pau Monné, Stefano Stabellini, Volodymyr Babchuk, Grygorii Strashko This patch introduces SCI driver to support for ARM EL3 Trusted Firmware-A (TF-A) which provides SCMI interface with multi-agent support, as shown below. +-----------------------------------------+ | | | EL3 TF-A SCMI | +-------+--+-------+--+-------+--+-------++ |shmem1 | |shmem0 | |shmem2 | |shmemX | +-----+-+ +---+---+ +--+----+ +---+---+ smc-id1 | | | | agent1 | | | | +-----v--------+---------+-----------+----+ | | | | | | | | | | +--------------+---------+-----------+----+ smc-id0 | smc-id2| smc-idX| agent0 | agent2 | agentX | | | | +----v---+ +--v-----+ +--v-----+ | | | | | | | Dom0 | | Dom1 | | DomX | | | | | | | | | | | | | +--------+ +--------+ +--------+ The EL3 SCMI multi-agent firmware is expected to provide SCMI SMC shared memory transport for every Agent in the system. The SCMI Agent transport channel defined by pair: - smc-id: SMC id used for Doorbell - shmem: shared memory for messages transfer, Xen page aligned. Shared memory is mapped with the following flags: MT_DEVICE_nGnRE. The following SCMI Agents are expected to be defined by SCMI FW to enable SCMI multi-agent functionality under Xen: - Xen management agent: trusted agents that accesses to the Base Protocol commands to configure agent specific permissions - OSPM VM agents: non-trusted agent, one for each Guest domain which is allowed direct HW access. At least one OSPM VM agent has to be provided by FW if HW is handled only by Dom0 or Driver Domain. The EL3 SCMI FW is expected to implement following Base protocol messages: - BASE_DISCOVER_AGENT (required for Xen and secondary agents without reg) - BASE_RESET_AGENT_CONFIGURATION (optional) - BASE_SET_DEVICE_PERMISSIONS (optional) The driver binds to the arm,scmi-smc management node inside a Xen-specific container compatible with xen,sci under /chosen/xen. Other SCMI nodes, including the host /firmware/scmi, are ignored by the mediator. Describe secondary agents as xen,scmi-agent children of an agents node inside the same container. The agents node uses #address-cells = <1> and `#size-cells = <0>`. Each child provides arm,smc-id and a single shmem phandle. The optional single-cell reg property supplies the SCMI agent ID (0..254). When reg is absent, Xen discovers the ID using BASE_DISCOVER_AGENT on that channel. Use agent@<id> for an explicit ID, and a node name without a unit address for discovery. Disabled agent nodes are ignored; at least one enabled secondary agent is required. Resolve discovered IDs before adding channels to the list, so duplicate IDs are rejected consistently, including clashes with Xen's management agent. Explicit IDs avoid mapping and probing secondary channels during initialization. Xen always discovers its own management agent ID. Keep shared-memory nodes under xen_scmi_config and leave the platform's existing /firmware/scmi and host OSPM shared-memory node unchanged. Channel nodes describe locations, rather than reserve memory: if firmware buffers occupy system RAM, the platform DT must reserve the whole firmware-owned range in /reserved-memory before Xen boots. A single reservation may cover multiple channels, including unused ones. SRAM outside system RAM does not require a RAM reservation. Channel nodes need not be moved into /reserved-memory. For example: / { #address-cells = <2>; #size-cells = <2>; scmi_shm_0: sram@47ff0000 { compatible = "arm,scmi-shmem"; reg = <0x0 0x47ff0000 0x0 0x1000>; }; firmware { scmi: scmi { compatible = "arm,scmi-smc"; arm,smc-id = <0x82000002>; #address-cells = <1>; #size-cells = <0>; shmem = <&scmi_shm_0>; }; }; chosen { #address-cells = <2>; #size-cells = <2>; ranges; xen { #address-cells = <2>; #size-cells = <2>; ranges; xen_scmi_config { compatible = "xen,sci"; #address-cells = <2>; #size-cells = <2>; ranges; xen,dom0-sci-agent-id = <0>; /* Xen management channel */ scmi_shm_1: sram@47ff1000 { compatible = "arm,scmi-shmem"; reg = <0x0 0x47ff1000 0x0 0x1000>; }; scmi_shm_2: sram@47ff2000 { compatible = "arm,scmi-shmem"; reg = <0x0 0x47ff2000 0x0 0x1000>; }; scmi_shm_3: sram@47ff3000 { compatible = "arm,scmi-shmem"; reg = <0x0 0x47ff3000 0x0 0x1000>; }; scmi_xen: scmi { compatible = "arm,scmi-smc"; arm,smc-id = <0x82000003>; #address-cells = <1>; #size-cells = <0>; #access-controller-cells = <1>; shmem = <&scmi_shm_1>; }; agents { #address-cells = <1>; #size-cells = <0>; agent@0 { compatible = "xen,scmi-agent"; reg = <0>; arm,smc-id = <0x82000002>; shmem = <&scmi_shm_0>; }; agent@2 { compatible = "xen,scmi-agent"; reg = <2>; arm,smc-id = <0x82000004>; shmem = <&scmi_shm_2>; }; agent@3 { compatible = "xen,scmi-agent"; reg = <3>; arm,smc-id = <0x82000005>; shmem = <&scmi_shm_3>; }; }; }; }; }; }; To discover a secondary agent ID instead, omit reg from its node: agent-domu { compatible = "xen,scmi-agent"; arm,smc-id = <0x82000004>; shmem = <&scmi_shm_2>; }; The agents child nodes replace the former scmi-secondary-agents list and `#scmi-secondary-agents-cells` property. The xen,sci subtree is omitted from Dom0's DT; the host OSPM channel is passed through when SCMI is enabled for Dom0. Full configuration and domain examples are included in the binding documentation and the hypervisor guide in patch 6. SMC ID Configuration and SCMI Connection Compatibility: The configuration allows the same device tree to work for both baremetal Linux and Linux Dom0. This is achieved because: - Baremetal Linux uses: func_id 0x82000002, scmi-shmem 0x47ff0000 - Dom0 Linux uses: func_id 0x82000002, scmi-shmem 0x47ff0000 - Xen management uses: func_id 0x82000003, scmi-shmem 0x47ff1000 This works because the privileged SCMI connection in EL3 firmware is not tied exclusively to func_id 0x82000002. The EL3 firmware supports multiple SCMI agents with different SMC IDs and shared memory regions. Each agent (Dom0 via 0x82000002, Xen via 0x82000003, other domains via additional func_ids) has an independent communication channel to the firmware. The key distinction is that Xen's management channel (0x82000003) is used for privileged operations like agent configuration and device permissions (BASE_SET_DEVICE_PERMISSIONS, BASE_RESET_AGENT_CONFIGURATION), while Dom0's channel (0x82000002) is used for standard SCMI protocol operations (power, clock, sensor management, etc.). The firmware enforces different permission levels for each agent based on their agent_id, not the SMC ID. Therefore, there is no conflict: Linux Dom0 retains its standard SCMI connection for hardware management, while Xen uses its separate privileged channel for mediating access between multiple domains. - It implements the SCI subsystem interface required for configuring and enabling SCMI functionality for Dom0/hwdom and Guest domains. To enable SCMI functionality for domain it has to be configured with unique supported SCMI Agent_id and use corresponding SCMI SMC shared memory transport [smc-id, shmem] defined for this SCMI Agent_id. - Once Xen domain is configured it can communicate with EL3 SCMI FW: -- zero-copy, the guest domain puts SCMI message in shmem; -- the guest triggers SMC exception with smc-id (doorbell); -- the Xen driver catches exception, do checks and synchronously forwards it to EL3 FW. - the Xen driver sends BASE_RESET_AGENT_CONFIGURATION message to Xen management agent channel on domain destroy event. This allows to reset resources used by domain and so implement use-case like domain reboot. Dom0 Enable SCMI SMC: - set xen,dom0-sci-agent-id=<agent_id> under the xen,sci container in the Host DT. If the property is absent, SCMI is disabled for Dom0 and all SCMI nodes are removed from the Dom0 DT. The driver updates the Dom0 DT SCMI node "arm,smc-id" value and fixes up the shmem node according to the assigned agent_id. - pass dom0=sci-agent-id=<agent_id> in Xen command line. if not provided SCMI will be disabled for Dom0 and all SCMI nodes removed from Dom0 DT. The driver updates Dom0 DT SCMI node "arm,smc-id" value and fix up shmem node according to assigned agent_id. Guest domains enable SCMI SMC: - xl.cfg: add configuration option as below arm_sci = "type=scmi_smc_multiagent,agent_id=2" - xl.cfg: enable access to the "arm,scmi-shmem" which should correspond assigned agent_id for the domain, for example: iomem = [ "47ff2,1@22001", ] - DT: add SCMI nodes to the Driver domain partial device tree as in the below example. The "arm,smc-id" should correspond assigned agent_id for the domain: passthrough { scmi_shm_0: sram@22001000 { compatible = "arm,scmi-shmem"; reg = <0x0 0x22001000 0x0 0x1000>; }; firmware { compatible = "simple-bus"; scmi: scmi { compatible = "arm,scmi-smc"; arm,smc-id = <0x82000004>; shmem = <&scmi_shm_0>; ... } } } SCMI "4.2.1.1 Device specific access control" The XEN SCI SCMI SMC multi-agent driver performs "access-controller" provider function in case EL3 SCMI FW implements SCMI "4.2.1.1 Device specific access control" and provides the BASE_SET_DEVICE_PERMISSIONS command to configure the devices that an agents have access to. The DT SCMI node should "#access-controller-cells=<1>" property and DT devices should be bound to the Xen SCMI. &i2c1 { access-controllers = <&scmi 0>; }; The Dom0 and dom0less domains DT devices will be processed automatically through sci_assign_dt_device() call, but to assign SCMI devices from toolstack the xl.cfg:"dtdev" property shall be used: dtdev = [ "/soc/i2c@e6508000", ] xl.cfg:dtdev will contain all nodes which are under SCMI management (not only those which are behind IOMMU). Additionally, this patch adds documentation for the pre-existing scmi-smc-passthrough command line option, which was previously undocumented. [0] https://developer.arm.com/documentation/den0056 [1] https://web.git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/firmware/arm,scmi.yaml [2] https://web.git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/access-controllers/access-controllers.yaml Signed-off-by: Grygorii Strashko <grygorii_strashko@epam.com> Signed-off-by: Oleksii Moisieiev <oleksii_moisieiev@epam.com> --- Changes in v14: - replace the flat secondary-agent list and its cell-count property with xen,scmi-agent child nodes under agents, using reg for an optional ID - scope secondary-agent parsing to the management channel's xen,sci container - resolve discovered IDs before insertion and reject duplicate or invalid IDs - validate shared-memory phandles, addresses and nonzero page-aligned sizes - retain shared-memory nodes under xen_scmi_config and document platform reservations separately, following the review discussion - update binding examples and commit description for the child-node format - match the const-qualified SCI domain_sanitise_config callback interface Changes in v13: - fix typo in commit description Changes in v12: - rebase to the latest staging - return EINVAL if xen,sci_type=scmi_smc_multiagent was set but CONFIG_SCMI_SMC_MA not set - put arm_sci_agent_id after v8r_el1_msa and before pad to preserve order. Reduced pad from uint16_t to uint8_t to fill the gap (was 2 bytes before adding arm_sci_agent_id and left only 1, so used uint8_t to preserve structure size) - fix scmi_handle_call to use arm_smccc_guest_smc helper function - update send_smc_message function to use new format of arm_smccc_1_1_smc - guard the Dom0 SCMI agent id setup in create_dom0() and MA related funcitons with CONFIG_SCMI_SMC_MA - fix the xen_scmi_config example in booting.txt, which placed properties after subnodes and was rejected by dtc with "Properties must precede subnodes" Changes in v10: - Fix tabs in MAINTAINERS file - remove duplicate SPDX tag from scmi-shmem.c - add cast to ARM_SMCCC_INVALID_PARAMETER to settle the sign since ARM_SMCCC_INVALID_PARAMETER is -3 which is part of the spec and resp is the default smccc call structure. - update free_channel_list. Add spinlock to avoid race condition and a comment with a description of the function work - preserve error of smc_create_channel in scmi_probe - check scmi shmem address alignment as wel as it is done for size - check for d->arch.sci_data != NULL in scmi_handle_call - use SCMI_SHMEM_MAPPED size for iomem_permit_access - change len type to unsigned in shmem_{get|put}_message - rename shmem_channel_is_free to shmem_channel_status - add comment about skipping message status when getting message response - Set correct agent_id ranges for dom0less and a toolstack. Agent_id 0 is not binded for dom0 so can be reused. Also mentioned that UINT8_MAX (255) is treated as invalid agent_id. - Split hypervisor and toolstack changes into separate commits - move init list and spin init after initial checks in probe call - fix typo in comments - clean resources when sci_register returns an error. Changes in v9: - sort and refactor MAINTAINERS enties - remove Spurious changes - add extra check to avoid ASSERT when calling unmap_channel_memory from assign device method - set correct tx flag to SCMI_BASE_AGENT_PERMISSIONS_RESET when freeing resources. Flag should be set to 1 according to the section 4.2.2.12 [0]. - fix dt node copmaring - moved channel->shmem check from ASSERT in unmap_memory_channel to "if" statement. This will prevent firing ASSERT if unmap_channel_memory was called twice on the same channel. Changes in v8: - update xen_scmi func_id in commit description - updated documentation with the new DT format - updated opt_dom0_scmi_agent_id setting to avoid it to be equal SCMI_AGENT_ID_INVALID. - changed SCMI_AGENT_ID_INVALID from 0xff to UINT8_MAX which makes code more clear showing that UINT8_MAX is theated like invalid agent_id and couldn't be used. Also excluded SCMI_AGENT_ID_INVALID from acceptable value range - remove outdated xen,config property ignore, added xen,sci compatible to skip_matches in handle_node - add documentation for pre-existing scmi-smc-passthrough command line option in alphabetically correct location (in 's' section) - add note to commit description about documentation for previously undocumented scmi-smc-passthrough - Fix SMC IDs in DT examples (Xen management uses 0x82000003, Dom0 uses 0x82000002) - Add explicit note explaining why Dom0 and Xen channels do not conflict - Document dom0less multi-agent configuration example (xen,sci_type / xen,sci-agent-id) - Add scmi_xen node to agent-discovery example with #scmi-secondary-agents-cells = 2 - Drop dom0=sci-agent-id command line handling; Dom0 SCMI is now enabled via xen,dom0-sci-agent-id in the xen,sci DT container - Refresh docs and examples to mention the DT property instead of the cmdline option Changes in v7: - rework scmi nodes for xen to match on compatible string instead of the direct path Changes in v6: - updated scmi-shmem to use io.h from generic location - update scmi_agent_id parameter to be provided inside dom0= parameter list and have the following format "dom0=sci-agent-id=0" This change was done as a response for Stefano comment and requires a lot of code changes, but produces much cleaner solution that's why I've added it to the code. - fix file comments and return codes - fix lenght checks in shmem_{get,put}_message to use offsetof - remove len member from scmi_channel structure as it is not used - set scmi-secondary-agents property to be mandatory since if no secondary agents were provided then there is no sence to enable scmi when no secondary agents are populated to the Domains - update documentation in booting.txt, added xen_scmi node to the example - adjust d->arch.sci_enabled value in scmi_domain_destroy - fix lock management in smc_create_channel call - avoid extra map_channel_memory command for Xen management channel because collect_agent_id call unmaps memory if DOMID_XEN is not set. So for Xen management channel we can init domain_id ad DOMID_XEN before calling collect_agent_id so memory shouldn't be unmapped. Changes in v5: - fix device-tree example format in booting.txt, added ";" after "}". - update define in scmi-proto.h - update define in scmi-shmem.h file - scmi_assign_device - do not ignore -EOPNOTSUPP return code of the do_smc_xfer - remove overwriting agent_channel->agent_id after SCMI_BASE_DISCOVER_AGENT call - add multi-agent files to the MAINTAINERS - add SCMI multi-agent description to the SUPPORT.md - handle ARM_SMCCC_INVALID_PARAMETER return code and return -EINVAL for smc call - updated collect_agents function. Set agent_id parameter as optional in scmi-secondary-agents device-tree property - introduce "#scmi-secondary-agents-cells" parameter to set if agent_id was provided - reanme xen,scmi-secondary-agents property to scmi-secondary-agents - move memcpu_toio/fromio for the generic place - update Xen to get management channel from /chosen/xen,config node - get hypervisor channnel from node instead of using hardcoded - update handling scmi and shmem nodes for the domain - Set multi-agent driver to support only Arm64 Changes in v4: - toolstack comments from Anthony PERARD - added dom0less support - added doc for "xen,scmi-secondary-agents" MAINTAINERS | 1 + SUPPORT.md | 11 + docs/misc/arm/device-tree/booting.txt | 195 +++++ xen/arch/arm/dom0less-build.c | 17 + xen/arch/arm/domain_build.c | 43 ++ xen/arch/arm/firmware/Kconfig | 12 + xen/arch/arm/firmware/Makefile | 1 + xen/arch/arm/firmware/scmi-proto.h | 164 ++++ xen/arch/arm/firmware/scmi-shmem.c | 118 +++ xen/arch/arm/firmware/scmi-shmem.h | 45 ++ xen/arch/arm/firmware/scmi-smc-multiagent.c | 792 ++++++++++++++++++++ xen/include/public/arch-arm.h | 5 +- 12 files changed, 1403 insertions(+), 1 deletion(-) create mode 100644 xen/arch/arm/firmware/scmi-proto.h create mode 100644 xen/arch/arm/firmware/scmi-shmem.c create mode 100644 xen/arch/arm/firmware/scmi-shmem.h create mode 100644 xen/arch/arm/firmware/scmi-smc-multiagent.c diff --git a/MAINTAINERS b/MAINTAINERS index c60afc93a5..cd6d1243bd 100644 --- a/MAINTAINERS +++ b/MAINTAINERS @@ -534,6 +534,7 @@ SCI MEDIATORS R: Oleksii Moisieiev <oleksii_moisieiev@epam.com> S: Supported F: xen/arch/arm/firmware/sci.c +F: xen/arch/arm/firmware/scmi-*.[ch] F: xen/arch/arm/include/asm/firmware/sci.h SEABIOS UPSTREAM diff --git a/SUPPORT.md b/SUPPORT.md index eb07332462..dc41a6b2f3 100644 --- a/SUPPORT.md +++ b/SUPPORT.md @@ -972,6 +972,17 @@ by hwdom. Some platforms use SCMI for access to system-level resources. Status: Supported +### Arm: SCMI SMC multi-agent support + +Enable support for the multi-agent configuration of the EL3 Firmware, which +allows Xen to provide an SCMI interface to the Domains. +Xen manages access permissions to the HW resources and provides an SCMI interface +to the Domains. Each Domain is represented as a separate Agent, which can +communicate with EL3 Firmware using a dedicated shared memory region, and +notifications are passed through by Xen. + + Status, ARM64: Tech Preview + ### ARM: Guest PSCI support Emulated PSCI interface exposed to guests. We support all mandatory diff --git a/docs/misc/arm/device-tree/booting.txt b/docs/misc/arm/device-tree/booting.txt index 3787d2e2b6..178f2493d8 100644 --- a/docs/misc/arm/device-tree/booting.txt +++ b/docs/misc/arm/device-tree/booting.txt @@ -331,6 +331,21 @@ with the following properties: Should be used together with scmi-smc-passthrough Xen command line option. + - "scmi_smc_multiagent" + + Enables ARM SCMI SMC multi-agent support for the guest by enabling SCMI over + SMC calls forwarding from domain to the EL3 firmware (like ARM + Trusted Firmware-A) with a multi SCMI OSPM agent support. + The SCMI agent_id should be specified for the guest with "xen,sci-agent-id" + property. + +- "xen,sci-agent-id" + + Specifies ARM SCMI agent id for the guest. This option is mandatory if the + SCMI SMC "scmi_smc_multiagent" support is enabled for the guest. The agent ids + of guest must be unique and in the range [0..254]. UINT8_MAX (255) is + treated as invalid. + - v8r_el1_msa A string property specifying whether, on Armv8-R systems at EL1, a domain @@ -847,3 +862,183 @@ The automatically allocated static shared memory will get mapped at 0x80000000 in DomU1 guest physical address space, and at 0x90000000 in DomU2 guest physical address space. DomU1 is explicitly defined as the owner domain, and DomU2 is the borrower domain. + +SCMI SMC multi-agent support +============================ + +For enabling the ARM SCMI SMC multi-agent support (enabled by CONFIG_SCMI_SMC_MA) +the Xen specific SCMI Agent's configuration shall be provided in the Host DT +according to the SCMI compliant EL3 Firmware specification with ARM SMC/HVC +transport. The SCMI configuration must live under the Xen SCMI container +"xen,sci" beneath "/chosen" (for example "/chosen/xen/xen_scmi_config/scmi"). The +Xen SCMI mediator will bind only to the "arm,scmi-smc" node that is a child of +this "xen,sci" container; any other "arm,scmi-smc" nodes (for example under +"/firmware") are ignored to avoid stealing the host's SCMI OSPM instance. + +- agents + + Required child of the xen,sci container. Describes the secondary SCMI + agents available for assignment to domains. Set #address-cells = <1> + and #size-cells = <0> in this node. At least one enabled child with + compatible = "xen,scmi-agent" is required. Each such child has: + + - reg: Optional single-cell SCMI agent ID in the range 0..254. When + present, name the node agent@<id>, with the unit address in hexadecimal. + When absent, use a name without a unit address; Xen discovers the ID + using BASE_DISCOVER_AGENT on that channel. A malformed reg is rejected. + - arm,smc-id: Required SMC function ID for this agent's doorbell. + - shmem: Required single phandle to an arm,scmi-shmem node. Its address + and nonzero size must be Xen page aligned. + + Agent IDs must be unique, including IDs returned by discovery, and must + not identify Xen's management agent. Children marked disabled are ignored. + The management arm,scmi-smc node remains a sibling of agents, directly + under xen,sci. + +- xen,dom0-sci-agent-id + + Optional. Specifies the Dom0/hwdom SCMI agent_id inside the ``xen,sci`` + container. When provided, Dom0 will be configured for SCMI multi-agent + support; when omitted, SCMI remains disabled for Dom0. The selected + agent's ``func_id`` and shared-memory region must match the host + ``/firmware/scmi`` channel passed to Dom0. + +The shared-memory nodes may remain inside xen_scmi_config, while the +platform's existing /firmware/scmi and its shared-memory node stay unchanged. +Shared-memory nodes describe channel locations; they do not reserve RAM. +If the channel buffers occupy system RAM, the platform DT must reserve the +whole firmware-owned range in /reserved-memory before Xen boots. One +reservation may cover several channels, including channels not listed in +this configuration. SRAM outside system RAM does not require a RAM +reservation. Moving the channel nodes into /reserved-memory is not required. + +As an example: + +/ { + #address-cells = <2>; + #size-cells = <2>; + + scmi_shm_0: sram@47ff0000 { + compatible = "arm,scmi-shmem"; + reg = <0x0 0x47ff0000 0x0 0x1000>; + }; + + firmware { + scmi: scmi { + compatible = "arm,scmi-smc"; + arm,smc-id = <0x82000002>; + #address-cells = <1>; + #size-cells = <0>; + shmem = <&scmi_shm_0>; + }; + }; + + chosen { + #address-cells = <2>; + #size-cells = <2>; + ranges; + + xen { + #address-cells = <2>; + #size-cells = <2>; + ranges; + + xen_scmi_config { + compatible = "xen,sci"; + #address-cells = <2>; + #size-cells = <2>; + ranges; + xen,dom0-sci-agent-id = <0>; /* dom0 agent id */ + + /* Xen management channel */ + scmi_shm_1: sram@47ff1000 { + compatible = "arm,scmi-shmem"; + reg = <0x0 0x47ff1000 0x0 0x1000>; + }; + + scmi_shm_2: sram@47ff2000 { + compatible = "arm,scmi-shmem"; + reg = <0x0 0x47ff2000 0x0 0x1000>; + }; + + scmi_shm_3: sram@47ff3000 { + compatible = "arm,scmi-shmem"; + reg = <0x0 0x47ff3000 0x0 0x1000>; + }; + + scmi_xen: scmi { + compatible = "arm,scmi-smc"; /* Xen management agent func_id */ + arm,smc-id = <0x82000003>; + #address-cells = <1>; + #size-cells = <0>; + #access-controller-cells = <1>; + shmem = <&scmi_shm_1>; /* Xen management agent shmem */ + }; + + agents { + #address-cells = <1>; + #size-cells = <0>; + + agent@0 { + compatible = "xen,scmi-agent"; + reg = <0>; + arm,smc-id = <0x82000002>; + shmem = <&scmi_shm_0>; + }; + + agent@2 { + compatible = "xen,scmi-agent"; + reg = <2>; + arm,smc-id = <0x82000004>; + shmem = <&scmi_shm_2>; + }; + + agent@3 { + compatible = "xen,scmi-agent"; + reg = <3>; + arm,smc-id = <0x82000005>; + shmem = <&scmi_shm_3>; + }; + }; + }; + }; + }; +}; + +This keeps the host OSPM channel (SMC 0x82000002, shmem 0x47ff0000) separate +from Xen's privileged management channel (SMC 0x82000003, shmem 0x47ff1000). +The xen,sci subtree is omitted from Dom0's device tree. + +To discover a secondary agent ID, omit reg from its node. For example, +replace only agent@2 above with: + +agent-domu { + compatible = "xen,scmi-agent"; + arm,smc-id = <0x82000004>; + shmem = <&scmi_shm_2>; +}; + +BASE_DISCOVER_AGENT is required on secondary channels only when reg is +omitted. Xen always uses it to discover its management agent ID. Providing +reg avoids mapping and probing the secondary channel during initialization. +The scmi-secondary-agents and #scmi-secondary-agents-cells properties used +by earlier versions are replaced by the agents child nodes. + +Dom0less example (multi-agent) +------------------------------- + +Use the same Xen SCMI configuration as above. Remove xen,dom0-sci-agent-id +if Dom0 does not use SCMI. Add xen,sci_type and xen,sci-agent-id to the +predefined guest's node under /chosen: + +chosen { + xen,domain@1 { + compatible = "xen,domain"; + xen,sci_type = "scmi_smc_multiagent"; + xen,sci-agent-id = <2>; + /* Additional domain properties (memory, cpus, kernels, etc.) */ + }; +}; + +The guest's partial DT and shared-memory mapping must use the SMC function +ID and shared-memory channel associated with agent 2 in the agents node. diff --git a/xen/arch/arm/dom0less-build.c b/xen/arch/arm/dom0less-build.c index 7bbb2eafb6..01c2a5ce52 100644 --- a/xen/arch/arm/dom0less-build.c +++ b/xen/arch/arm/dom0less-build.c @@ -293,6 +293,23 @@ static int __init domu_dt_sci_parse(struct dt_device_node *node, d_cfg->arch.arm_sci_type = XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC; } + else if ( !strcmp(sci_type, "scmi_smc_multiagent") ) + { + uint32_t agent_id = 0; + + if ( !IS_ENABLED(CONFIG_SCMI_SMC_MA) ) + { + printk(XENLOG_ERR "xen,sci_type=scmi_smc_multiagent requested, but CONFIG_SCMI_SMC_MA not set\n"); + return -EINVAL; + } + + if ( !dt_property_read_u32(node, "xen,sci-agent-id", &agent_id) || + agent_id >= UINT8_MAX ) + return -EINVAL; + + d_cfg->arch.arm_sci_type = XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC_MA; + d_cfg->arch.arm_sci_agent_id = agent_id; + } else { printk(XENLOG_ERR "xen,sci_type in not valid (%s) for domain %s\n", diff --git a/xen/arch/arm/domain_build.c b/xen/arch/arm/domain_build.c index 72f6f7539a..5d42cdf6af 100644 --- a/xen/arch/arm/domain_build.c +++ b/xen/arch/arm/domain_build.c @@ -88,6 +88,39 @@ int __init parse_arch_dom0_param(const char *s, const char *e) return -EINVAL; } +#ifdef CONFIG_SCMI_SMC_MA +/* SCMI agent ID for dom0 obtained from xen,sci container */ +#define SCMI_AGENT_ID_INVALID UINT8_MAX + +static uint8_t __init get_dom0_scmi_agent_id(void) +{ + const struct dt_device_node *config_node; + u32 val; + const struct dt_property *prop; + + config_node = dt_find_compatible_node(NULL, NULL, "xen,sci"); + if ( !config_node ) + return SCMI_AGENT_ID_INVALID; + + prop = dt_find_property(config_node, "xen,dom0-sci-agent-id", NULL); + if ( !prop ) + return SCMI_AGENT_ID_INVALID; + + if ( !dt_property_read_u32(config_node, "xen,dom0-sci-agent-id", &val) ) + return SCMI_AGENT_ID_INVALID; + + if ( val >= SCMI_AGENT_ID_INVALID ) + { + printk(XENLOG_WARNING + "Invalid xen,dom0-sci-agent-id=%u, SCMI disabled for Dom0\n", + val); + return SCMI_AGENT_ID_INVALID; + } + + return val; +} +#endif /* CONFIG_SCMI_SMC_MA */ + /* Override macros from asm/page.h to make them work with mfn_t */ #undef virt_to_mfn #define virt_to_mfn(va) _mfn(__virt_to_mfn(va)) @@ -1480,6 +1513,7 @@ static int __init handle_node(struct domain *d, struct kernel_info *kinfo, DT_MATCH_TYPE("memory"), /* The memory mapped timer is not supported by Xen. */ DT_MATCH_COMPATIBLE("arm,armv7-timer-mem"), + DT_MATCH_COMPATIBLE("xen,sci"), { /* sentinel */ }, }; static const struct dt_device_match timer_matches[] __initconst = @@ -1975,6 +2009,15 @@ void __init create_dom0(void) dom0_cfg.arch.tee_type = tee_get_type(); dom0_cfg.max_vcpus = dom0_max_vcpus(); +#ifdef CONFIG_SCMI_SMC_MA + /* Set up SCMI agent ID if provided in the xen,sci container */ + dom0_cfg.arch.arm_sci_agent_id = get_dom0_scmi_agent_id(); + dom0_cfg.arch.arm_sci_type = (dom0_cfg.arch.arm_sci_agent_id != + SCMI_AGENT_ID_INVALID) ? + XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC_MA : + XEN_DOMCTL_CONFIG_ARM_SCI_NONE; +#endif + if ( iommu_enabled ) dom0_cfg.flags |= XEN_DOMCTL_CDF_iommu; diff --git a/xen/arch/arm/firmware/Kconfig b/xen/arch/arm/firmware/Kconfig index 5c5f0880c4..972cd9b173 100644 --- a/xen/arch/arm/firmware/Kconfig +++ b/xen/arch/arm/firmware/Kconfig @@ -29,6 +29,18 @@ config SCMI_SMC driver domain. Use with EL3 firmware which supports only single SCMI OSPM agent. +config SCMI_SMC_MA + bool "Enable ARM SCMI SMC multi-agent driver" + depends on ARM_64 + select ARM_SCI + help + Enables SCMI SMC/HVC multi-agent in XEN to pass SCMI requests from Domains + to EL3 firmware (TF-A) which supports multi-agent feature. + This feature allows to enable SCMI per Domain using unique SCMI agent_id, + so Domain is identified by EL3 firmware as an SCMI Agent and can access + allowed platform resources through dedicated SMC/HVC Shared memory based + transport. + endchoice endmenu diff --git a/xen/arch/arm/firmware/Makefile b/xen/arch/arm/firmware/Makefile index 71bdefc24a..37927e690e 100644 --- a/xen/arch/arm/firmware/Makefile +++ b/xen/arch/arm/firmware/Makefile @@ -1,2 +1,3 @@ obj-$(CONFIG_ARM_SCI) += sci.o obj-$(CONFIG_SCMI_SMC) += scmi-smc.o +obj-$(CONFIG_SCMI_SMC_MA) += scmi-shmem.o scmi-smc-multiagent.o diff --git a/xen/arch/arm/firmware/scmi-proto.h b/xen/arch/arm/firmware/scmi-proto.h new file mode 100644 index 0000000000..49f63cfc0a --- /dev/null +++ b/xen/arch/arm/firmware/scmi-proto.h @@ -0,0 +1,164 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * Arm System Control and Management Interface definitions + * Version 3.0 (DEN0056C) + * + * Copyright (c) 2025 EPAM Systems + */ + +#ifndef ARM_FIRMWARE_SCMI_PROTO_H_ +#define ARM_FIRMWARE_SCMI_PROTO_H_ + +#include <xen/stdint.h> + +#define SCMI_SHORT_NAME_MAX_SIZE 16 + +/* SCMI status codes. See section 4.1.4 */ +#define SCMI_SUCCESS 0 +#define SCMI_NOT_SUPPORTED (-1) +#define SCMI_INVALID_PARAMETERS (-2) +#define SCMI_DENIED (-3) +#define SCMI_NOT_FOUND (-4) +#define SCMI_OUT_OF_RANGE (-5) +#define SCMI_BUSY (-6) +#define SCMI_COMMS_ERROR (-7) +#define SCMI_GENERIC_ERROR (-8) +#define SCMI_HARDWARE_ERROR (-9) +#define SCMI_PROTOCOL_ERROR (-10) + +/* Protocol IDs */ +#define SCMI_BASE_PROTOCOL 0x10 + +/* Base protocol message IDs */ +#define SCMI_BASE_PROTOCOL_VERSION 0x0 +#define SCMI_BASE_PROTOCOL_ATTIBUTES 0x1 +#define SCMI_BASE_PROTOCOL_MESSAGE_ATTRIBUTES 0x2 +#define SCMI_BASE_DISCOVER_AGENT 0x7 +#define SCMI_BASE_SET_DEVICE_PERMISSIONS 0x9 +#define SCMI_BASE_RESET_AGENT_CONFIGURATION 0xB + +typedef struct scmi_msg_header { + uint8_t id; + uint8_t type; + uint8_t protocol; + uint32_t status; +} scmi_msg_header_t; + +/* Table 2 Message header format */ +#define SCMI_HDR_ID GENMASK(7, 0) +#define SCMI_HDR_TYPE GENMASK(9, 8) +#define SCMI_HDR_PROTO GENMASK(17, 10) + +#define SCMI_FIELD_GET(_mask, _reg) \ + ((typeof(_mask))(((_reg) & (_mask)) >> (ffs64(_mask) - 1))) +#define SCMI_FIELD_PREP(_mask, _val) \ + (((typeof(_mask))(_val) << (ffs64(_mask) - 1)) & (_mask)) + +static inline uint32_t pack_scmi_header(scmi_msg_header_t *hdr) +{ + return SCMI_FIELD_PREP(SCMI_HDR_ID, hdr->id) | + SCMI_FIELD_PREP(SCMI_HDR_TYPE, hdr->type) | + SCMI_FIELD_PREP(SCMI_HDR_PROTO, hdr->protocol); +} + +static inline void unpack_scmi_header(uint32_t msg_hdr, scmi_msg_header_t *hdr) +{ + hdr->id = SCMI_FIELD_GET(SCMI_HDR_ID, msg_hdr); + hdr->type = SCMI_FIELD_GET(SCMI_HDR_TYPE, msg_hdr); + hdr->protocol = SCMI_FIELD_GET(SCMI_HDR_PROTO, msg_hdr); +} + +static inline int scmi_to_xen_errno(int scmi_status) +{ + if ( scmi_status == SCMI_SUCCESS ) + return 0; + + switch ( scmi_status ) + { + case SCMI_NOT_SUPPORTED: + return -EOPNOTSUPP; + case SCMI_INVALID_PARAMETERS: + return -EINVAL; + case SCMI_DENIED: + return -EACCES; + case SCMI_NOT_FOUND: + return -ENOENT; + case SCMI_OUT_OF_RANGE: + return -ERANGE; + case SCMI_BUSY: + return -EBUSY; + case SCMI_COMMS_ERROR: + return -ENOTCONN; + case SCMI_GENERIC_ERROR: + return -EIO; + case SCMI_HARDWARE_ERROR: + return -ENXIO; + case SCMI_PROTOCOL_ERROR: + return -EBADMSG; + default: + return -EINVAL; + } +} + +/* PROTOCOL_VERSION */ +#define SCMI_VERSION_MINOR GENMASK(15, 0) +#define SCMI_VERSION_MAJOR GENMASK(31, 16) + +struct scmi_msg_prot_version_p2a { + uint32_t version; +} __packed; + +/* BASE PROTOCOL_ATTRIBUTES */ +#define SCMI_BASE_ATTR_NUM_PROTO GENMASK(7, 0) +#define SCMI_BASE_ATTR_NUM_AGENT GENMASK(15, 8) + +struct scmi_msg_base_attributes_p2a { + uint32_t attributes; +} __packed; + +/* + * BASE_DISCOVER_AGENT + */ +#define SCMI_BASE_AGENT_ID_OWN 0xFFFFFFFF + +struct scmi_msg_base_discover_agent_a2p { + uint32_t agent_id; +} __packed; + +struct scmi_msg_base_discover_agent_p2a { + uint32_t agent_id; + char name[SCMI_SHORT_NAME_MAX_SIZE]; +} __packed; + +/* + * BASE_SET_DEVICE_PERMISSIONS + */ +#define SCMI_BASE_DEVICE_ACCESS_ALLOW BIT(0, UL) + +struct scmi_msg_base_set_device_permissions_a2p { + uint32_t agent_id; + uint32_t device_id; + uint32_t flags; +} __packed; + +/* + * BASE_RESET_AGENT_CONFIGURATION + */ +#define SCMI_BASE_AGENT_PERMISSIONS_RESET BIT(0, UL) + +struct scmi_msg_base_reset_agent_cfg_a2p { + uint32_t agent_id; + uint32_t flags; +} __packed; + +#endif /* ARM_FIRMWARE_SCMI_PROTO_H_ */ + +/* + * Local variables: + * mode: C + * c-file-style: "BSD" + * c-basic-offset: 4 + * tab-width: 4 + * indent-tabs-mode: nil + * End: + */ diff --git a/xen/arch/arm/firmware/scmi-shmem.c b/xen/arch/arm/firmware/scmi-shmem.c new file mode 100644 index 0000000000..e36745a85e --- /dev/null +++ b/xen/arch/arm/firmware/scmi-shmem.c @@ -0,0 +1,118 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * SMC/HVC shmem transport implementation used by + * SCI SCMI multi-agent driver. + * + * Oleksii Moisieiev <oleksii_moisieiev@epam.com> + * Copyright (c) 2025 EPAM Systems + */ + +#include <xen/err.h> +#include <xen/io.h> +#include <asm/io.h> + +#include "scmi-proto.h" +#include "scmi-shmem.h" + +static inline int +shmem_channel_status(const volatile struct scmi_shared_mem __iomem *shmem) +{ + return (readl(&shmem->channel_status) & + SCMI_SHMEM_CHAN_STAT_CHANNEL_FREE) ? 0 : -EBUSY; +} + +int shmem_put_message(volatile struct scmi_shared_mem __iomem *shmem, + scmi_msg_header_t *hdr, void *data, unsigned int len) +{ + int ret; + + if ( (len + offsetof(struct scmi_shared_mem, msg_payload)) > + SCMI_SHMEM_MAPPED_SIZE ) + { + printk(XENLOG_ERR "scmi: Wrong size of smc message. Data is invalid\n"); + return -EINVAL; + } + + ret = shmem_channel_status(shmem); + if ( ret ) + return ret; + + writel_relaxed(0x0, &shmem->channel_status); + /* Writing 0x0 right now, but "shmem"_FLAG_INTR_ENABLED can be set */ + writel_relaxed(0x0, &shmem->flags); + writel_relaxed(sizeof(shmem->msg_header) + len, &shmem->length); + writel(pack_scmi_header(hdr), &shmem->msg_header); + + if ( len > 0 && data ) + memcpy_toio(shmem->msg_payload, data, len); + + return 0; +} + +int shmem_get_response(const volatile struct scmi_shared_mem __iomem *shmem, + scmi_msg_header_t *hdr, void *data, unsigned int len) +{ + int recv_len; + int ret; + /* + * First word of msg_payload carries the returned status; exclude it from + * recv_len so only the protocol payload is copied back to the caller. + */ + int pad = sizeof(hdr->status); + + if ( len >= SCMI_SHMEM_MAPPED_SIZE - + offsetof(struct scmi_shared_mem, msg_payload) ) + { + printk(XENLOG_ERR + "scmi: Wrong size of input smc message. Data may be invalid\n"); + return -EINVAL; + } + + ret = shmem_channel_status(shmem); + if ( ret ) + return ret; + + recv_len = readl(&shmem->length) - sizeof(shmem->msg_header); + + if ( recv_len < 0 ) + { + printk(XENLOG_ERR + "scmi: Wrong size of smc message. Data may be invalid\n"); + return -EINVAL; + } + + unpack_scmi_header(readl(&shmem->msg_header), hdr); + + hdr->status = readl(&shmem->msg_payload); + recv_len = recv_len > pad ? recv_len - pad : 0; + + ret = scmi_to_xen_errno(hdr->status); + if ( ret ) + { + printk(XENLOG_DEBUG "scmi: Error received: %d\n", ret); + return ret; + } + + if ( recv_len > len ) + { + printk(XENLOG_ERR + "scmi: Not enough buffer for message %d, expecting %d\n", + recv_len, len); + return -EINVAL; + } + + if ( recv_len > 0 ) + memcpy_fromio(data, shmem->msg_payload + pad, recv_len); + + return 0; +} + +/* + * Local variables: + * mode: C + * c-file-style: "BSD" + * c-basic-offset: 4 + * tab-width: 4 + * indent-tabs-mode: nil + * End: + */ diff --git a/xen/arch/arm/firmware/scmi-shmem.h b/xen/arch/arm/firmware/scmi-shmem.h new file mode 100644 index 0000000000..722263aa77 --- /dev/null +++ b/xen/arch/arm/firmware/scmi-shmem.h @@ -0,0 +1,45 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * Arm System Control and Management Interface definitions + * Version 3.0 (DEN0056C) + * Shared Memory based Transport + * + * Copyright (c) 2024 EPAM Systems + */ + +#ifndef ARM_FIRMWARE_SCMI_SHMEM_H_ +#define ARM_FIRMWARE_SCMI_SHMEM_H_ + +#include <xen/stdint.h> + +#define SCMI_SHMEM_CHAN_STAT_CHANNEL_FREE BIT(0, UL) +#define SCMI_SHMEM_CHAN_STAT_CHANNEL_ERROR BIT(1, UL) + +struct scmi_shared_mem { + uint32_t reserved; + uint32_t channel_status; + uint32_t reserved1[2]; + uint32_t flags; + uint32_t length; + uint32_t msg_header; + uint8_t msg_payload[]; +}; + +#define SCMI_SHMEM_MAPPED_SIZE PAGE_SIZE + +int shmem_put_message(volatile struct scmi_shared_mem __iomem *shmem, + scmi_msg_header_t *hdr, void *data, unsigned int len); + +int shmem_get_response(const volatile struct scmi_shared_mem __iomem *shmem, + scmi_msg_header_t *hdr, void *data, unsigned int len); +#endif /* ARM_FIRMWARE_SCMI_SHMEM_H_ */ + +/* + * Local variables: + * mode: C + * c-file-style: "BSD" + * c-basic-offset: 4 + * tab-width: 4 + * indent-tabs-mode: nil + * End: + */ diff --git a/xen/arch/arm/firmware/scmi-smc-multiagent.c b/xen/arch/arm/firmware/scmi-smc-multiagent.c new file mode 100644 index 0000000000..6c46f6cd36 --- /dev/null +++ b/xen/arch/arm/firmware/scmi-smc-multiagent.c @@ -0,0 +1,792 @@ +/* SPDX-License-Identifier: GPL-2.0-only */ +/* + * SCI SCMI multi-agent driver, using SMC/HVC shmem as transport. + * + * Oleksii Moisieiev <oleksii_moisieiev@epam.com> + * Copyright (c) 2025 EPAM Systems + */ + +#include <xen/acpi.h> + +#include <xen/device_tree.h> +#include <xen/init.h> +#include <xen/iocap.h> +#include <xen/err.h> +#include <xen/libfdt/libfdt.h> +#include <xen/string.h> +#include <xen/param.h> +#include <xen/sched.h> +#include <xen/vmap.h> + +#include <asm/firmware/sci.h> +#include <asm/smccc.h> + +#include "scmi-proto.h" +#include "scmi-shmem.h" + +struct scmi_channel { + uint32_t agent_id; + uint32_t func_id; + domid_t domain_id; + uint64_t paddr; + struct scmi_shared_mem __iomem *shmem; + spinlock_t lock; + struct list_head list; +}; + +struct scmi_data { + struct list_head channel_list; + spinlock_t channel_list_lock; + uint32_t func_id; + bool initialized; + uint32_t shmem_phandle; + uint32_t hyp_channel_agent_id; + struct dt_device_node *dt_dev; +}; + +static struct scmi_data scmi_data; + +static bool scmi_is_under_xen_sci(const struct dt_device_node *node) +{ + const struct dt_device_node *p; + + for ( p = node->parent; p; p = p->parent ) + if ( dt_device_is_compatible(p, "xen,sci") ) + return true; + + return false; +} + +static int send_smc_message(struct scmi_channel *chan_info, + scmi_msg_header_t *hdr, void *data, int len) +{ + struct arm_smccc_res resp; + int ret; + + ret = shmem_put_message(chan_info->shmem, hdr, data, len); + if ( ret ) + return ret; + + resp = arm_smccc_1_1_smc(chan_info->func_id, 0, 0, 0, 0, 0, 0, 0); + + if ( resp.a0 == (unsigned long)ARM_SMCCC_INVALID_PARAMETER ) + return -EINVAL; + + if ( resp.a0 ) + return -EOPNOTSUPP; + + return 0; +} + +static int do_smc_xfer(struct scmi_channel *chan_info, scmi_msg_header_t *hdr, + void *tx_data, int tx_size, void *rx_data, int rx_size) +{ + int ret = 0; + + ASSERT(chan_info && chan_info->shmem); + + if ( !hdr ) + return -EINVAL; + + spin_lock(&chan_info->lock); + + printk(XENLOG_DEBUG + "scmi: agent_id = %d msg_id = %x type = %d, proto = %x\n", + chan_info->agent_id, hdr->id, hdr->type, hdr->protocol); + + ret = send_smc_message(chan_info, hdr, tx_data, tx_size); + if ( ret ) + goto clean; + + ret = shmem_get_response(chan_info->shmem, hdr, rx_data, rx_size); + +clean: + printk(XENLOG_DEBUG + "scmi: get smc response agent_id = %d msg_id = %x proto = %x res=%d\n", + chan_info->agent_id, hdr->id, hdr->protocol, ret); + + spin_unlock(&chan_info->lock); + + return ret; +} + +static struct scmi_channel *get_channel_by_id(uint32_t agent_id) +{ + struct scmi_channel *curr; + bool found = false; + + spin_lock(&scmi_data.channel_list_lock); + list_for_each_entry(curr, &scmi_data.channel_list, list) + { + if ( curr->agent_id == agent_id ) + { + found = true; + break; + } + } + + spin_unlock(&scmi_data.channel_list_lock); + if ( found ) + return curr; + + return NULL; +} + +static struct scmi_channel *acquire_scmi_channel(struct domain *d, + uint32_t agent_id) +{ + struct scmi_channel *curr; + struct scmi_channel *ret = ERR_PTR(-ENOENT); + + spin_lock(&scmi_data.channel_list_lock); + list_for_each_entry(curr, &scmi_data.channel_list, list) + { + if ( curr->agent_id == agent_id ) + { + if ( curr->domain_id != DOMID_INVALID ) + { + ret = ERR_PTR(-EEXIST); + break; + } + + curr->domain_id = d->domain_id; + ret = curr; + break; + } + } + + spin_unlock(&scmi_data.channel_list_lock); + + return ret; +} + +static void relinquish_scmi_channel(struct scmi_channel *channel) +{ + ASSERT(channel != NULL); + + spin_lock(&scmi_data.channel_list_lock); + channel->domain_id = DOMID_INVALID; + spin_unlock(&scmi_data.channel_list_lock); +} + +static int map_channel_memory(struct scmi_channel *channel) +{ + ASSERT(channel && channel->paddr); + channel->shmem = ioremap_nocache(channel->paddr, SCMI_SHMEM_MAPPED_SIZE); + if ( !channel->shmem ) + return -ENOMEM; + + channel->shmem->channel_status = SCMI_SHMEM_CHAN_STAT_CHANNEL_FREE; + printk(XENLOG_DEBUG "scmi: Got shmem %lx after vmap %p\n", channel->paddr, + channel->shmem); + + return 0; +} + +static void unmap_channel_memory(struct scmi_channel *channel) +{ + ASSERT(channel); + + if ( !channel->shmem ) + return; + + iounmap(channel->shmem); + channel->shmem = NULL; +} + +static struct scmi_channel *smc_create_channel(uint32_t agent_id, + uint32_t func_id, uint64_t addr) +{ + struct scmi_channel *channel, *curr; + + spin_lock(&scmi_data.channel_list_lock); + + /* Check if channel already exists while holding the lock */ + list_for_each_entry(curr, &scmi_data.channel_list, list) + { + if ( curr->agent_id == agent_id ) + { + spin_unlock(&scmi_data.channel_list_lock); + return ERR_PTR(-EEXIST); + } + } + + channel = xmalloc(struct scmi_channel); + if ( !channel ) + { + spin_unlock(&scmi_data.channel_list_lock); + return ERR_PTR(-ENOMEM); + } + + spin_lock_init(&channel->lock); + channel->agent_id = agent_id; + channel->func_id = func_id; + channel->domain_id = DOMID_INVALID; + channel->shmem = NULL; + channel->paddr = addr; + list_add_tail(&channel->list, &scmi_data.channel_list); + + spin_unlock(&scmi_data.channel_list_lock); + return channel; +} + +static void free_channel_list(void) +{ + struct scmi_channel *curr, *_curr; + /* + * Called only on the __init error path, before any runtime users exist, + * so no other thread is iterating channel_list. Keep the lock for the + * whole drain for clarity and to avoid misleading readers about possible + * concurrent access. + */ + spin_lock(&scmi_data.channel_list_lock); + list_for_each_entry_safe(curr, _curr, &scmi_data.channel_list, list) + { + list_del(&curr->list); + xfree(curr); + } + spin_unlock(&scmi_data.channel_list_lock); +} + +static int __init +scmi_dt_read_channel_addr(struct dt_device_node *scmi_node, u64 *addr, + u64 *size) +{ + struct dt_device_node *shmem_node; + uint32_t phandle; + int ret; + + ret = dt_property_read_variable_u32_array(scmi_node, "shmem", &phandle, + 1, 1); + if ( ret < 0 ) + return ret; + + shmem_node = dt_find_node_by_phandle(phandle); + if ( !shmem_node || + !dt_device_is_compatible(shmem_node, "arm,scmi-shmem") ) + return -EINVAL; + + ret = dt_device_get_address(shmem_node, 0, addr, size); + if ( ret ) + return ret; + + if ( !*size || !IS_ALIGNED(*size, SCMI_SHMEM_MAPPED_SIZE) || + !IS_ALIGNED(*addr, SCMI_SHMEM_MAPPED_SIZE) ) + { + printk(XENLOG_ERR "scmi: invalid shmem range for %s\n", + scmi_node->full_name); + return -EINVAL; + } + + return 0; +} + +/* + * Handle Dom0 SCMI specific DT nodes + * + * Make a decision on copying SCMI specific nodes into Dom0 device tree. + * For SCMI multi-agent case: + * - shmem nodes will not be copied and generated instead if SCMI + * is enabled for Dom0 + * - scmi node will be copied if SCMI is enabled for Dom0 + */ +static bool scmi_dt_handle_node(struct domain *d, struct dt_device_node *node) +{ + static const struct dt_device_match shmem_matches[] __initconst = { + DT_MATCH_COMPATIBLE("arm,scmi-shmem"), + { /* sentinel */ }, + }; + static const struct dt_device_match scmi_matches[] __initconst = { + DT_MATCH_PATH("/firmware/scmi"), + { /* sentinel */ }, + }; + + if ( !scmi_data.initialized ) + return false; + + /* skip scmi shmem node for dom0 if scmi not enabled */ + if ( dt_match_node(shmem_matches, node) && !sci_domain_is_enabled(d) ) + { + dt_dprintk(" Skip scmi shmem node\n"); + return true; + } + + /* drop scmi if not enabled */ + if ( dt_match_node(scmi_matches, node) && !sci_domain_is_enabled(d) ) + { + dt_dprintk(" Skip scmi node\n"); + return true; + } + + return false; +} + +static int scmi_assign_device(uint32_t agent_id, uint32_t device_id, + uint32_t flags) +{ + struct scmi_msg_base_set_device_permissions_a2p tx; + struct scmi_channel *channel; + scmi_msg_header_t hdr; + + channel = get_channel_by_id(scmi_data.hyp_channel_agent_id); + if ( !channel ) + return -EINVAL; + + hdr.id = SCMI_BASE_SET_DEVICE_PERMISSIONS; + hdr.type = 0; + hdr.protocol = SCMI_BASE_PROTOCOL; + + tx.agent_id = agent_id; + tx.device_id = device_id; + tx.flags = flags; + + return do_smc_xfer(channel, &hdr, &tx, sizeof(tx), NULL, 0); +} + +static int scmi_dt_assign_device(struct domain *d, + struct dt_phandle_args *ac_spec) +{ + struct scmi_channel *agent_channel; + uint32_t scmi_device_id = ac_spec->args[0]; + int ret; + + if ( !d->arch.sci_data ) + return 0; + + /* The access-controllers is specified for DT dev, but it's not a SCMI */ + if ( !scmi_data.dt_dev || + !dt_node_path_is_equal(ac_spec->np, scmi_data.dt_dev->full_name) ) + return 0; + + agent_channel = d->arch.sci_data; + + spin_lock(&agent_channel->lock); + + ret = scmi_assign_device(agent_channel->agent_id, scmi_device_id, + SCMI_BASE_DEVICE_ACCESS_ALLOW); + if ( ret ) + { + printk(XENLOG_ERR + "scmi: could not assign dev for %pd agent:%d dev_id:%u (%d)", + d, agent_channel->agent_id, scmi_device_id, ret); + } + + spin_unlock(&agent_channel->lock); + return ret; +} + +static int collect_agent_id(struct scmi_channel *agent_channel) +{ + int ret; + scmi_msg_header_t hdr; + struct scmi_msg_base_discover_agent_p2a da_rx; + struct scmi_msg_base_discover_agent_a2p da_tx; + + ret = map_channel_memory(agent_channel); + if ( ret ) + return ret; + + hdr.id = SCMI_BASE_DISCOVER_AGENT; + hdr.type = 0; + hdr.protocol = SCMI_BASE_PROTOCOL; + + da_tx.agent_id = agent_channel->agent_id; + + ret = do_smc_xfer(agent_channel, &hdr, &da_tx, sizeof(da_tx), &da_rx, + sizeof(da_rx)); + if ( agent_channel->domain_id != DOMID_XEN ) + unmap_channel_memory(agent_channel); + if ( ret ) + return ret; + + printk(XENLOG_DEBUG "id=0x%x name=%s\n", da_rx.agent_id, da_rx.name); + agent_channel->agent_id = da_rx.agent_id; + return 0; +} + +static int __init collect_agents(struct dt_device_node *scmi_node) +{ + struct dt_device_node *agents, *node; + unsigned int count = 0; + + /* Use the configuration containing this management channel. */ + for ( agents = scmi_node->parent; agents; agents = agents->parent ) + if ( dt_device_is_compatible(agents, "xen,sci") ) + break; + + ASSERT(agents); + + dt_for_each_child_node(agents, node) + if ( dt_node_name_is_equal(node, "agents") ) + break; + + if ( !node ) + { + printk(XENLOG_ERR "scmi: missing agents node\n"); + return -EINVAL; + } + agents = node; + + if ( dt_child_n_addr_cells(agents) != 1 || + dt_child_n_size_cells(agents) != 0 ) + return -EINVAL; + + dt_for_each_child_node(agents, node) + { + struct scmi_channel agent = { + .agent_id = SCMI_BASE_AGENT_ID_OWN, + .domain_id = DOMID_INVALID, + }; + struct scmi_channel *channel; + bool discover; + uint32_t len; + u64 size; + int ret; + + if ( !dt_device_is_compatible(node, "xen,scmi-agent") || + !dt_device_is_available(node) ) + continue; + + discover = !dt_find_property(node, "reg", &len); + if ( !discover ) + { + if ( len != sizeof(uint32_t) || + !dt_property_read_u32(node, "reg", &agent.agent_id) ) + return -EINVAL; + } + + if ( !dt_property_read_u32(node, "arm,smc-id", &agent.func_id) ) + return -EINVAL; + + ret = scmi_dt_read_channel_addr(node, &agent.paddr, &size); + if ( ret ) + return ret; + + if ( discover ) + { + /* Resolve the ID before inserting it into the channel list. */ + spin_lock_init(&agent.lock); + ret = collect_agent_id(&agent); + if ( ret ) + return ret; + } + + /* Guest configurations encode agent IDs in an eight-bit field. */ + if ( agent.agent_id >= UINT8_MAX ) + return -EINVAL; + + channel = smc_create_channel(agent.agent_id, agent.func_id, + agent.paddr); + if ( IS_ERR(channel) ) + { + printk(XENLOG_ERR "scmi: Could not create channel for %s: %ld\n", + node->full_name, PTR_ERR(channel)); + return PTR_ERR(channel); + } + + count++; + printk(XENLOG_DEBUG "scmi: Agent %u SMC %X addr %lx\n", + agent.agent_id, agent.func_id, (unsigned long)agent.paddr); + } + + return count ? 0 : -EINVAL; +} + +static int scmi_domain_init(struct domain *d, + struct xen_domctl_createdomain *config) +{ + struct scmi_channel *channel; + int ret; + + if ( !scmi_data.initialized ) + return 0; + + /* + * SCMI support is configured via: + * - For dom0: xen,dom0-sci-agent-id property under the xen,sci container + * - For dom0less: xen,sci-agent-id in the domain node + * The config->arch.arm_sci_type and config->arch.arm_sci_agent_id + * are already set by domain_build.c or dom0less-build.c + */ + + if ( config->arch.arm_sci_type == XEN_DOMCTL_CONFIG_ARM_SCI_NONE ) + return 0; + + channel = acquire_scmi_channel(d, config->arch.arm_sci_agent_id); + if ( IS_ERR(channel) ) + { + printk(XENLOG_ERR + "scmi: Failed to acquire SCMI channel for agent_id %u: %ld\n", + config->arch.arm_sci_agent_id, PTR_ERR(channel)); + return PTR_ERR(channel); + } + + printk(XENLOG_INFO + "scmi: Acquire channel id = 0x%x, domain_id = %d paddr = 0x%lx\n", + channel->agent_id, channel->domain_id, channel->paddr); + + /* + * Dom0 (if present) needs to have an access to the guest memory range + * to satisfy iomem_access_permitted() check in XEN_DOMCTL_iomem_permission + * domctl. + */ + if ( hardware_domain && !is_hardware_domain(d) ) + { + ret = iomem_permit_access(hardware_domain, paddr_to_pfn(channel->paddr), + paddr_to_pfn(channel->paddr + + SCMI_SHMEM_MAPPED_SIZE - 1)); + if ( ret ) + goto error; + } + + d->arch.sci_data = channel; + d->arch.sci_enabled = true; + + return 0; + +error: + relinquish_scmi_channel(channel); + return ret; +} + +int scmi_domain_sanitise_config(const struct xen_domctl_createdomain *config) +{ + if ( config->arch.arm_sci_type != XEN_DOMCTL_CONFIG_ARM_SCI_NONE && + config->arch.arm_sci_type != XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC_MA ) + { + dprintk(XENLOG_INFO, "scmi: Unsupported ARM_SCI type\n"); + return -EINVAL; + } + + return 0; +} + +static int scmi_relinquish_resources(struct domain *d) +{ + int ret; + struct scmi_channel *channel, *agent_channel; + scmi_msg_header_t hdr; + struct scmi_msg_base_reset_agent_cfg_a2p tx; + + if ( !d->arch.sci_data ) + return 0; + + agent_channel = d->arch.sci_data; + + spin_lock(&agent_channel->lock); + tx.agent_id = agent_channel->agent_id; + spin_unlock(&agent_channel->lock); + + channel = get_channel_by_id(scmi_data.hyp_channel_agent_id); + if ( !channel ) + { + printk(XENLOG_ERR + "scmi: Unable to get Hypervisor scmi channel for domain %d\n", + d->domain_id); + return -EINVAL; + } + + hdr.id = SCMI_BASE_RESET_AGENT_CONFIGURATION; + hdr.type = 0; + hdr.protocol = SCMI_BASE_PROTOCOL; + + tx.flags = SCMI_BASE_AGENT_PERMISSIONS_RESET; + + ret = do_smc_xfer(channel, &hdr, &tx, sizeof(tx), NULL, 0); + if ( ret == -EOPNOTSUPP ) + return 0; + + return ret; +} + +static void scmi_domain_destroy(struct domain *d) +{ + struct scmi_channel *channel; + + if ( !d->arch.sci_data ) + return; + + channel = d->arch.sci_data; + spin_lock(&channel->lock); + + relinquish_scmi_channel(channel); + printk(XENLOG_DEBUG "scmi: Free domain %d\n", d->domain_id); + + d->arch.sci_data = NULL; + d->arch.sci_enabled = false; + + spin_unlock(&channel->lock); +} + +static bool scmi_handle_call(struct cpu_user_regs *regs) +{ + uint32_t fid = (uint32_t)get_user_reg(regs, 0); + struct scmi_channel *agent_channel; + struct domain *d = current->domain; + bool res = false; + + if ( (!sci_domain_is_enabled(d)) || (!d->arch.sci_data) ) + return false; + + agent_channel = d->arch.sci_data; + spin_lock(&agent_channel->lock); + + if ( agent_channel->func_id != fid ) + { + res = false; + goto unlock; + } + + arm_smccc_guest_smc(regs); + res = true; +unlock: + spin_unlock(&agent_channel->lock); + + return res; +} + +static const struct sci_mediator_ops scmi_ops = { + .domain_init = scmi_domain_init, + .domain_destroy = scmi_domain_destroy, + .relinquish_resources = scmi_relinquish_resources, + .handle_call = scmi_handle_call, + .dom0_dt_handle_node = scmi_dt_handle_node, + .domain_sanitise_config = scmi_domain_sanitise_config, + .assign_dt_device = scmi_dt_assign_device, +}; + +static int __init scmi_check_smccc_ver(void) +{ + if ( smccc_ver < ARM_SMCCC_VERSION_1_1 ) + { + printk(XENLOG_WARNING + "scmi: No SMCCC 1.1 support, SCMI calls forwarding disabled\n"); + return -ENOSYS; + } + + return 0; +} + +static int __init scmi_dt_hyp_channel_read(struct dt_device_node *scmi_node, + struct scmi_data *scmi_data, + u64 *addr) +{ + u64 size; + + if ( !dt_property_read_u32(scmi_node, "arm,smc-id", &scmi_data->func_id) ) + { + printk(XENLOG_ERR "scmi: unable to read smc-id from DT\n"); + return -ENOENT; + } + + return scmi_dt_read_channel_addr(scmi_node, addr, &size); +} + +static __init int scmi_probe(struct dt_device_node *scmi_node, const void *data) +{ + u64 addr; + int ret; + struct scmi_channel *channel; + unsigned int n_agents; + scmi_msg_header_t hdr; + struct scmi_msg_base_attributes_p2a rx; + + ASSERT(scmi_node != NULL); + + /* + * Only bind to the SCMI node provided by Xen under the xen,sci container + * (e.g. /chosen/xen/xen_scmi_config/scmi). This avoids binding to firmware + * SCMI nodes that belong to the host OSPM and keeps the mediator scoped to + * Xen-provided configuration only. + */ + if ( !scmi_is_under_xen_sci(scmi_node) ) + return -ENODEV; + + if ( !acpi_disabled ) + { + printk(XENLOG_WARNING "scmi: is not supported when using ACPI\n"); + return -EINVAL; + } + + ret = scmi_check_smccc_ver(); + if ( ret ) + return ret; + + ret = scmi_dt_hyp_channel_read(scmi_node, &scmi_data, &addr); + if ( ret ) + return ret; + + INIT_LIST_HEAD(&scmi_data.channel_list); + spin_lock_init(&scmi_data.channel_list_lock); + + scmi_data.dt_dev = scmi_node; + + channel = smc_create_channel(SCMI_BASE_AGENT_ID_OWN, scmi_data.func_id, addr); + if ( IS_ERR(channel) ) + { + ret = PTR_ERR(channel); + goto out; + } + + /* Mark as Xen management channel before collecting agent ID */ + channel->domain_id = DOMID_XEN; + + /* Request agent id for Xen management channel */ + ret = collect_agent_id(channel); + if ( ret ) + goto error; + + /* Save the agent id for Xen management channel */ + scmi_data.hyp_channel_agent_id = channel->agent_id; + + hdr.id = SCMI_BASE_PROTOCOL_ATTIBUTES; + hdr.type = 0; + hdr.protocol = SCMI_BASE_PROTOCOL; + + ret = do_smc_xfer(channel, &hdr, NULL, 0, &rx, sizeof(rx)); + if ( ret ) + goto error; + + n_agents = SCMI_FIELD_GET(SCMI_BASE_ATTR_NUM_AGENT, rx.attributes); + printk(XENLOG_DEBUG "scmi: Got agent count %d\n", n_agents); + ret = collect_agents(scmi_node); + if ( ret ) + goto error; + + ret = sci_register(&scmi_ops); + if ( ret ) + { + printk(XENLOG_ERR "SCMI: mediator already registered (ret = %d)\n", + ret); + goto error; + } + + scmi_data.initialized = true; + goto out; + +error: + unmap_channel_memory(channel); + free_channel_list(); +out: + return ret; +} + +static const struct dt_device_match scmi_smc_match[] __initconst = { + DT_MATCH_COMPATIBLE("arm,scmi-smc"), + { /* sentinel */ }, +}; + +DT_DEVICE_START(scmi_smc_ma, "SCMI SMC MEDIATOR", DEVICE_FIRMWARE) + .dt_match = scmi_smc_match, + .init = scmi_probe, +DT_DEVICE_END + +/* + * Local variables: + * mode: C + * c-file-style: "BSD" + * c-basic-offset: 4 + * tab-width: 4 + * indent-tabs-mode: nil + * End: + */ diff --git a/xen/include/public/arch-arm.h b/xen/include/public/arch-arm.h index 14d51e0746..722c2d2e3a 100644 --- a/xen/include/public/arch-arm.h +++ b/xen/include/public/arch-arm.h @@ -329,6 +329,7 @@ DEFINE_XEN_GUEST_HANDLE(vcpu_guest_context_t); #define XEN_DOMCTL_CONFIG_ARM_SCI_NONE 0 #define XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC 1 +#define XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC_MA 2 #define XEN_DOMCTL_CONFIG_ARM_V8R_EL1_MSA_NONE 0 #define XEN_DOMCTL_CONFIG_ARM_V8R_EL1_MSA_PMSA 1 @@ -347,7 +348,9 @@ struct xen_arch_domainconfig { uint8_t arm_sci_type; /* IN */ uint8_t v8r_el1_msa; - uint16_t pad; + /* IN */ + uint8_t arm_sci_agent_id; + uint8_t pad; }; #endif /* __XEN__ || __XEN_TOOLS__ */ -- 2.43.0 ^ permalink raw reply related [flat|nested] 10+ messages in thread
* Re: [PATCH v14 4/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent driver 2026-10-20 8:43 ` [PATCH v14 4/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent driver Oleksii Moisieiev @ 2026-10-23 1:56 ` Volodymyr Babchuk 2026-10-26 15:39 ` Oleksii Moisieiev 0 siblings, 1 reply; 10+ messages in thread From: Volodymyr Babchuk @ 2026-10-23 1:56 UTC (permalink / raw) To: Oleksii Moisieiev Cc: xen-devel@lists.xenproject.org, Andrew Cooper, Anthony PERARD, Bertrand Marquis, Jan Beulich, Juergen Gross, Julien Grall, Michal Orzel, Roger Pau Monné, Stefano Stabellini, Grygorii Strashko Hi Oleksii, Oleksii Moisieiev <Oleksii_Moisieiev@epam.com> writes: > This patch introduces SCI driver to support for ARM EL3 Trusted Firmware-A > (TF-A) which provides SCMI interface with multi-agent support, as shown > below. > > +-----------------------------------------+ > | | > | EL3 TF-A SCMI | > +-------+--+-------+--+-------+--+-------++ > |shmem1 | |shmem0 | |shmem2 | |shmemX | > +-----+-+ +---+---+ +--+----+ +---+---+ > smc-id1 | | | | > agent1 | | | | > +-----v--------+---------+-----------+----+ > | | | | | > | | | | | > +--------------+---------+-----------+----+ > smc-id0 | smc-id2| smc-idX| > agent0 | agent2 | agentX | > | | | > +----v---+ +--v-----+ +--v-----+ > | | | | | | > | Dom0 | | Dom1 | | DomX | > | | | | | | > | | | | | | > +--------+ +--------+ +--------+ > > The EL3 SCMI multi-agent firmware is expected to provide SCMI SMC shared > memory transport for every Agent in the system. > > The SCMI Agent transport channel defined by pair: > - smc-id: SMC id used for Doorbell > - shmem: shared memory for messages transfer, Xen page > aligned. Shared memory is mapped with the following flags: > MT_DEVICE_nGnRE. > > The following SCMI Agents are expected to be defined by SCMI FW to enable SCMI > multi-agent functionality under Xen: > - Xen management agent: trusted agents that accesses to the Base Protocol > commands to configure agent specific permissions > - OSPM VM agents: non-trusted agent, one for each Guest domain which is > allowed direct HW access. At least one OSPM VM agent has to be provided > by FW if HW is handled only by Dom0 or Driver Domain. > > The EL3 SCMI FW is expected to implement following Base protocol messages: > - BASE_DISCOVER_AGENT (required for Xen and secondary agents without reg) > - BASE_RESET_AGENT_CONFIGURATION (optional) > - BASE_SET_DEVICE_PERMISSIONS (optional) > > The driver binds to the arm,scmi-smc management node inside a Xen-specific > container compatible with xen,sci under /chosen/xen. Other SCMI nodes, > including the host /firmware/scmi, are ignored by the mediator. > > Describe secondary agents as xen,scmi-agent children of an agents node > inside the same container. The agents node uses #address-cells = <1> and > `#size-cells = <0>`. Each child provides arm,smc-id and a single shmem phandle. > The optional single-cell reg property supplies the SCMI agent ID (0..254). > When reg is absent, Xen discovers the ID using BASE_DISCOVER_AGENT on that > channel. Use agent@<id> for an explicit ID, and a node name without a unit > address for discovery. Disabled agent nodes are ignored; at least one > enabled secondary agent is required. > > Resolve discovered IDs before adding channels to the list, so duplicate > IDs are rejected consistently, including clashes with Xen's management > agent. Explicit IDs avoid mapping and probing secondary channels during > initialization. Xen always discovers its own management agent ID. > > Keep shared-memory nodes under xen_scmi_config and leave the platform's > existing /firmware/scmi and host OSPM shared-memory node unchanged. Channel > nodes describe locations, rather than reserve memory: if firmware buffers > occupy system RAM, the platform DT must reserve the whole firmware-owned > range in /reserved-memory before Xen boots. A single reservation may cover > multiple channels, including unused ones. SRAM outside system RAM does > not require a RAM reservation. Channel nodes need not be moved into > /reserved-memory. > > For example: > > / { > #address-cells = <2>; > #size-cells = <2>; > > scmi_shm_0: sram@47ff0000 { > compatible = "arm,scmi-shmem"; > reg = <0x0 0x47ff0000 0x0 0x1000>; > }; > > firmware { > scmi: scmi { > compatible = "arm,scmi-smc"; > arm,smc-id = <0x82000002>; > #address-cells = <1>; > #size-cells = <0>; > shmem = <&scmi_shm_0>; > }; > }; > > chosen { > #address-cells = <2>; > #size-cells = <2>; > ranges; > > xen { > #address-cells = <2>; > #size-cells = <2>; > ranges; > > xen_scmi_config { > compatible = "xen,sci"; > #address-cells = <2>; > #size-cells = <2>; > ranges; > xen,dom0-sci-agent-id = <0>; > > /* Xen management channel */ > scmi_shm_1: sram@47ff1000 { > compatible = "arm,scmi-shmem"; > reg = <0x0 0x47ff1000 0x0 0x1000>; > }; > > scmi_shm_2: sram@47ff2000 { > compatible = "arm,scmi-shmem"; > reg = <0x0 0x47ff2000 0x0 0x1000>; > }; > > scmi_shm_3: sram@47ff3000 { > compatible = "arm,scmi-shmem"; > reg = <0x0 0x47ff3000 0x0 0x1000>; > }; > > scmi_xen: scmi { > compatible = "arm,scmi-smc"; > arm,smc-id = <0x82000003>; > #address-cells = <1>; > #size-cells = <0>; > #access-controller-cells = <1>; > shmem = <&scmi_shm_1>; > }; > > agents { > #address-cells = <1>; > #size-cells = <0>; > > agent@0 { > compatible = "xen,scmi-agent"; > reg = <0>; > arm,smc-id = <0x82000002>; > shmem = <&scmi_shm_0>; > }; > > agent@2 { > compatible = "xen,scmi-agent"; > reg = <2>; > arm,smc-id = <0x82000004>; > shmem = <&scmi_shm_2>; > }; > > agent@3 { > compatible = "xen,scmi-agent"; > reg = <3>; > arm,smc-id = <0x82000005>; > shmem = <&scmi_shm_3>; > }; > }; > }; > }; > }; > }; > > To discover a secondary agent ID instead, omit reg from its node: > > agent-domu { > compatible = "xen,scmi-agent"; > arm,smc-id = <0x82000004>; > shmem = <&scmi_shm_2>; > }; > > The agents child nodes replace the former scmi-secondary-agents list and > `#scmi-secondary-agents-cells` property. The xen,sci subtree is omitted from > Dom0's DT; the host OSPM channel is passed through when SCMI is enabled > for Dom0. Full configuration and domain examples are included in the > binding documentation and the hypervisor guide in patch 6. > > SMC ID Configuration and SCMI Connection Compatibility: > > The configuration allows the same device tree to work for both baremetal > Linux and Linux Dom0. This is achieved because: > > - Baremetal Linux uses: func_id 0x82000002, scmi-shmem 0x47ff0000 > - Dom0 Linux uses: func_id 0x82000002, scmi-shmem 0x47ff0000 > - Xen management uses: func_id 0x82000003, scmi-shmem 0x47ff1000 > > This works because the privileged SCMI connection in EL3 firmware is not > tied exclusively to func_id 0x82000002. The EL3 firmware supports multiple > SCMI agents with different SMC IDs and shared memory regions. Each agent > (Dom0 via 0x82000002, Xen via 0x82000003, other domains via additional > func_ids) has an independent communication channel to the firmware. > > The key distinction is that Xen's management channel (0x82000003) is used > for privileged operations like agent configuration and device permissions > (BASE_SET_DEVICE_PERMISSIONS, BASE_RESET_AGENT_CONFIGURATION), while Dom0's > channel (0x82000002) is used for standard SCMI protocol operations (power, > clock, sensor management, etc.). The firmware enforces different permission > levels for each agent based on their agent_id, not the SMC ID. > > Therefore, there is no conflict: Linux Dom0 retains its standard SCMI > connection for hardware management, while Xen uses its separate privileged > channel for mediating access between multiple domains. > > - It implements the SCI subsystem interface required for configuring and > enabling SCMI functionality for Dom0/hwdom and Guest domains. To enable > SCMI functionality for domain it has to be configured with unique supported > SCMI Agent_id and use corresponding SCMI SMC shared memory transport > [smc-id, shmem] defined for this SCMI Agent_id. > - Once Xen domain is configured it can communicate with EL3 SCMI FW: > -- zero-copy, the guest domain puts SCMI message in shmem; > -- the guest triggers SMC exception with smc-id (doorbell); > -- the Xen driver catches exception, do checks and synchronously forwards > it to EL3 FW. > - the Xen driver sends BASE_RESET_AGENT_CONFIGURATION message to Xen > management agent channel on domain destroy event. This allows to reset > resources used by domain and so implement use-case like domain reboot. > > Dom0 Enable SCMI SMC: > - set xen,dom0-sci-agent-id=<agent_id> under the xen,sci container in > the Host DT. If the property is absent, SCMI is disabled for Dom0 > and all SCMI nodes are removed from the Dom0 DT. The driver updates > the Dom0 DT SCMI node "arm,smc-id" value and fixes up the shmem > node according to the assigned agent_id. > > - pass dom0=sci-agent-id=<agent_id> in Xen command line. if not provided > SCMI will be disabled for Dom0 and all SCMI nodes removed from Dom0 DT. > The driver updates Dom0 DT SCMI node "arm,smc-id" value and fix up shmem > node according to assigned agent_id. > > Guest domains enable SCMI SMC: > - xl.cfg: add configuration option as below > > arm_sci = "type=scmi_smc_multiagent,agent_id=2" > > - xl.cfg: enable access to the "arm,scmi-shmem" which should > correspond assigned agent_id for the domain, for example: > > iomem = [ > "47ff2,1@22001", > ] > > - DT: add SCMI nodes to the Driver domain partial device tree as in the > below example. The "arm,smc-id" should correspond assigned agent_id > for the domain: > > passthrough { > scmi_shm_0: sram@22001000 { > compatible = "arm,scmi-shmem"; > reg = <0x0 0x22001000 0x0 0x1000>; > }; > > firmware { > compatible = "simple-bus"; > scmi: scmi { > compatible = "arm,scmi-smc"; > arm,smc-id = <0x82000004>; > shmem = <&scmi_shm_0>; > ... > } > } > } > > SCMI "4.2.1.1 Device specific access control" > > The XEN SCI SCMI SMC multi-agent driver performs "access-controller" > provider function in case EL3 SCMI FW implements SCMI "4.2.1.1 Device > specific access control" and provides the BASE_SET_DEVICE_PERMISSIONS > command to configure the devices that an agents have access to. > The DT SCMI node should "#access-controller-cells=<1>" property and DT > devices should be bound to the Xen SCMI. > > &i2c1 { > access-controllers = <&scmi 0>; > }; > > The Dom0 and dom0less domains DT devices will be processed > automatically through sci_assign_dt_device() call, but to assign SCMI > devices from toolstack the xl.cfg:"dtdev" property > shall be used: > > dtdev = [ > "/soc/i2c@e6508000", > ] > > xl.cfg:dtdev will contain all nodes which are under SCMI > management (not only those which are behind IOMMU). > > Additionally, this patch adds documentation for the pre-existing > scmi-smc-passthrough command line option, which was previously > undocumented. > > [0] https://developer.arm.com/documentation/den0056 > [1] https://web.git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/firmware/arm,scmi.yaml > [2] https://web.git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/access-controllers/access-controllers.yaml > > Signed-off-by: Grygorii Strashko <grygorii_strashko@epam.com> > Signed-off-by: Oleksii Moisieiev <oleksii_moisieiev@epam.com> > --- > > Changes in v14: > - replace the flat secondary-agent list and its cell-count property with > xen,scmi-agent child nodes under agents, using reg for an optional ID > - scope secondary-agent parsing to the management channel's xen,sci container > - resolve discovered IDs before insertion and reject duplicate or invalid IDs > - validate shared-memory phandles, addresses and nonzero page-aligned sizes > - retain shared-memory nodes under xen_scmi_config and document platform > reservations separately, following the review discussion > - update binding examples and commit description for the child-node format > - match the const-qualified SCI domain_sanitise_config callback interface > > Changes in v13: > - fix typo in commit description > > Changes in v12: > - rebase to the latest staging > - return EINVAL if xen,sci_type=scmi_smc_multiagent was set but > CONFIG_SCMI_SMC_MA not set > - put arm_sci_agent_id after v8r_el1_msa and before pad to preserve > order. Reduced pad from uint16_t to uint8_t to fill the gap (was 2 > bytes before adding arm_sci_agent_id and left only 1, so used uint8_t > to preserve structure size) > - fix scmi_handle_call to use arm_smccc_guest_smc helper function > - update send_smc_message function to use new format of > arm_smccc_1_1_smc > - guard the Dom0 SCMI agent id setup in create_dom0() and MA related > funcitons with CONFIG_SCMI_SMC_MA > - fix the xen_scmi_config example in booting.txt, which placed properties > after subnodes and was rejected by dtc with "Properties must precede > subnodes" > > Changes in v10: > - Fix tabs in MAINTAINERS file > - remove duplicate SPDX tag from scmi-shmem.c > - add cast to ARM_SMCCC_INVALID_PARAMETER to settle the sign since > ARM_SMCCC_INVALID_PARAMETER is -3 which is part of the spec and resp > is the default smccc call structure. > - update free_channel_list. Add spinlock to avoid race condition and > a comment with a description of the function work > - preserve error of smc_create_channel in scmi_probe > - check scmi shmem address alignment as wel as it is done for size > - check for d->arch.sci_data != NULL in scmi_handle_call > - use SCMI_SHMEM_MAPPED size for iomem_permit_access > - change len type to unsigned in shmem_{get|put}_message > - rename shmem_channel_is_free to shmem_channel_status > - add comment about skipping message status when getting message response > - Set correct agent_id ranges for dom0less and a toolstack. Agent_id 0 > is not binded for dom0 so can be reused. Also mentioned that > UINT8_MAX (255) is treated as invalid agent_id. > - Split hypervisor and toolstack changes into separate commits > - move init list and spin init after initial checks in probe call > - fix typo in comments > - clean resources when sci_register returns an error. > > Changes in v9: > - sort and refactor MAINTAINERS enties > - remove Spurious changes > - add extra check to avoid ASSERT when calling unmap_channel_memory > from assign device method > - set correct tx flag to SCMI_BASE_AGENT_PERMISSIONS_RESET when > freeing resources. Flag should be set to 1 according to the > section 4.2.2.12 [0]. > - fix dt node copmaring > - moved channel->shmem check from ASSERT in unmap_memory_channel to > "if" statement. This will prevent firing ASSERT if > unmap_channel_memory was called twice on the same channel. > > Changes in v8: > - update xen_scmi func_id in commit description > - updated documentation with the new DT format > - updated opt_dom0_scmi_agent_id setting to avoid it to be equal > SCMI_AGENT_ID_INVALID. > - changed SCMI_AGENT_ID_INVALID from 0xff to UINT8_MAX which makes > code more clear showing that UINT8_MAX is theated like invalid > agent_id and couldn't be used. Also excluded SCMI_AGENT_ID_INVALID > from acceptable value range > - remove outdated xen,config property ignore, added xen,sci compatible > to skip_matches in handle_node > - add documentation for pre-existing scmi-smc-passthrough command line > option in alphabetically correct location (in 's' section) > - add note to commit description about documentation for previously > undocumented scmi-smc-passthrough > - Fix SMC IDs in DT examples (Xen management uses 0x82000003, Dom0 uses 0x82000002) > - Add explicit note explaining why Dom0 and Xen channels do not conflict > - Document dom0less multi-agent configuration example (xen,sci_type / xen,sci-agent-id) > - Add scmi_xen node to agent-discovery example with #scmi-secondary-agents-cells = 2 > - Drop dom0=sci-agent-id command line handling; Dom0 SCMI is now enabled via > xen,dom0-sci-agent-id in the xen,sci DT container > - Refresh docs and examples to mention the DT property instead of the cmdline option > > Changes in v7: > - rework scmi nodes for xen to match on compatible string instead of > the direct path > > Changes in v6: > - updated scmi-shmem to use io.h from generic location > - update scmi_agent_id parameter to be provided inside dom0= parameter > list and have the following format "dom0=sci-agent-id=0" > This change was done as a response for Stefano comment and > requires a lot of code changes, but produces much cleaner solution > that's why I've added it to the code. > - fix file comments and return codes > - fix lenght checks in shmem_{get,put}_message to use offsetof > - remove len member from scmi_channel structure as it is not used > - set scmi-secondary-agents property to be mandatory since if no > secondary agents were provided then there is no sence to enable scmi > when no secondary agents are populated to the Domains > - update documentation in booting.txt, added xen_scmi node to the > example > - adjust d->arch.sci_enabled value in scmi_domain_destroy > - fix lock management in smc_create_channel call > - avoid extra map_channel_memory command for Xen management channel > because collect_agent_id call unmaps memory if DOMID_XEN is not > set. So for Xen management channel we can init domain_id ad DOMID_XEN > before calling collect_agent_id so memory shouldn't be unmapped. > > Changes in v5: > - fix device-tree example format in booting.txt, added ";" after "}". > - update define in scmi-proto.h > - update define in scmi-shmem.h file > - scmi_assign_device - do not ignore -EOPNOTSUPP return > code of the do_smc_xfer > - remove overwriting agent_channel->agent_id after > SCMI_BASE_DISCOVER_AGENT call > - add multi-agent files to the MAINTAINERS > - add SCMI multi-agent description to the SUPPORT.md > - handle ARM_SMCCC_INVALID_PARAMETER return code and return -EINVAL > for smc call > - updated collect_agents function. Set agent_id parameter as optional > in scmi-secondary-agents device-tree property > - introduce "#scmi-secondary-agents-cells" parameter to set if > agent_id was provided > - reanme xen,scmi-secondary-agents property to scmi-secondary-agents > - move memcpu_toio/fromio for the generic place > - update Xen to get management channel from /chosen/xen,config node > - get hypervisor channnel from node instead of using hardcoded > - update handling scmi and shmem nodes for the domain > - Set multi-agent driver to support only Arm64 > > Changes in v4: > - toolstack comments from Anthony PERARD > - added dom0less support > - added doc for "xen,scmi-secondary-agents" > > MAINTAINERS | 1 + > SUPPORT.md | 11 + > docs/misc/arm/device-tree/booting.txt | 195 +++++ > xen/arch/arm/dom0less-build.c | 17 + > xen/arch/arm/domain_build.c | 43 ++ > xen/arch/arm/firmware/Kconfig | 12 + > xen/arch/arm/firmware/Makefile | 1 + > xen/arch/arm/firmware/scmi-proto.h | 164 ++++ > xen/arch/arm/firmware/scmi-shmem.c | 118 +++ > xen/arch/arm/firmware/scmi-shmem.h | 45 ++ > xen/arch/arm/firmware/scmi-smc-multiagent.c | 792 ++++++++++++++++++++ > xen/include/public/arch-arm.h | 5 +- > 12 files changed, 1403 insertions(+), 1 deletion(-) > create mode 100644 xen/arch/arm/firmware/scmi-proto.h > create mode 100644 xen/arch/arm/firmware/scmi-shmem.c > create mode 100644 xen/arch/arm/firmware/scmi-shmem.h > create mode 100644 xen/arch/arm/firmware/scmi-smc-multiagent.c > > diff --git a/MAINTAINERS b/MAINTAINERS > index c60afc93a5..cd6d1243bd 100644 > --- a/MAINTAINERS > +++ b/MAINTAINERS > @@ -534,6 +534,7 @@ SCI MEDIATORS > R: Oleksii Moisieiev <oleksii_moisieiev@epam.com> > S: Supported > F: xen/arch/arm/firmware/sci.c > +F: xen/arch/arm/firmware/scmi-*.[ch] > F: xen/arch/arm/include/asm/firmware/sci.h > > SEABIOS UPSTREAM > diff --git a/SUPPORT.md b/SUPPORT.md > index eb07332462..dc41a6b2f3 100644 > --- a/SUPPORT.md > +++ b/SUPPORT.md > @@ -972,6 +972,17 @@ by hwdom. Some platforms use SCMI for access to system-level resources. > > Status: Supported > > +### Arm: SCMI SMC multi-agent support > + > +Enable support for the multi-agent configuration of the EL3 Firmware, which > +allows Xen to provide an SCMI interface to the Domains. > +Xen manages access permissions to the HW resources and provides an SCMI interface > +to the Domains. Each Domain is represented as a separate Agent, which can > +communicate with EL3 Firmware using a dedicated shared memory region, and > +notifications are passed through by Xen. > + > + Status, ARM64: Tech Preview > + > ### ARM: Guest PSCI support > > Emulated PSCI interface exposed to guests. We support all mandatory > diff --git a/docs/misc/arm/device-tree/booting.txt b/docs/misc/arm/device-tree/booting.txt > index 3787d2e2b6..178f2493d8 100644 > --- a/docs/misc/arm/device-tree/booting.txt > +++ b/docs/misc/arm/device-tree/booting.txt > @@ -331,6 +331,21 @@ with the following properties: > Should be used together with scmi-smc-passthrough Xen command line > option. > > + - "scmi_smc_multiagent" > + > + Enables ARM SCMI SMC multi-agent support for the guest by enabling SCMI over > + SMC calls forwarding from domain to the EL3 firmware (like ARM > + Trusted Firmware-A) with a multi SCMI OSPM agent support. > + The SCMI agent_id should be specified for the guest with "xen,sci-agent-id" > + property. > + > +- "xen,sci-agent-id" > + > + Specifies ARM SCMI agent id for the guest. This option is mandatory if the > + SCMI SMC "scmi_smc_multiagent" support is enabled for the guest. The agent ids > + of guest must be unique and in the range [0..254]. UINT8_MAX (255) is > + treated as invalid. > + > - v8r_el1_msa > > A string property specifying whether, on Armv8-R systems at EL1, a domain > @@ -847,3 +862,183 @@ The automatically allocated static shared memory will get mapped at > 0x80000000 in DomU1 guest physical address space, and at 0x90000000 in DomU2 > guest physical address space. DomU1 is explicitly defined as the owner domain, > and DomU2 is the borrower domain. > + > +SCMI SMC multi-agent support > +============================ > + > +For enabling the ARM SCMI SMC multi-agent support (enabled by CONFIG_SCMI_SMC_MA) > +the Xen specific SCMI Agent's configuration shall be provided in the Host DT > +according to the SCMI compliant EL3 Firmware specification with ARM SMC/HVC > +transport. The SCMI configuration must live under the Xen SCMI container > +"xen,sci" beneath "/chosen" (for example "/chosen/xen/xen_scmi_config/scmi"). The > +Xen SCMI mediator will bind only to the "arm,scmi-smc" node that is a child of > +this "xen,sci" container; any other "arm,scmi-smc" nodes (for example under > +"/firmware") are ignored to avoid stealing the host's SCMI OSPM instance. > + > +- agents > + > + Required child of the xen,sci container. Describes the secondary SCMI > + agents available for assignment to domains. Set #address-cells = <1> > + and #size-cells = <0> in this node. At least one enabled child with > + compatible = "xen,scmi-agent" is required. Each such child has: > + > + - reg: Optional single-cell SCMI agent ID in the range 0..254. When > + present, name the node agent@<id>, with the unit address in hexadecimal. > + When absent, use a name without a unit address; Xen discovers the ID > + using BASE_DISCOVER_AGENT on that channel. A malformed reg is rejected. > + - arm,smc-id: Required SMC function ID for this agent's doorbell. > + - shmem: Required single phandle to an arm,scmi-shmem node. Its address > + and nonzero size must be Xen page aligned. > + > + Agent IDs must be unique, including IDs returned by discovery, and must > + not identify Xen's management agent. Children marked disabled are ignored. > + The management arm,scmi-smc node remains a sibling of agents, directly > + under xen,sci. > + > +- xen,dom0-sci-agent-id > + > + Optional. Specifies the Dom0/hwdom SCMI agent_id inside the ``xen,sci`` > + container. When provided, Dom0 will be configured for SCMI multi-agent > + support; when omitted, SCMI remains disabled for Dom0. The selected > + agent's ``func_id`` and shared-memory region must match the host > + ``/firmware/scmi`` channel passed to Dom0. > + > +The shared-memory nodes may remain inside xen_scmi_config, while the > +platform's existing /firmware/scmi and its shared-memory node stay unchanged. > +Shared-memory nodes describe channel locations; they do not reserve RAM. > +If the channel buffers occupy system RAM, the platform DT must reserve the > +whole firmware-owned range in /reserved-memory before Xen boots. One > +reservation may cover several channels, including channels not listed in > +this configuration. SRAM outside system RAM does not require a RAM > +reservation. Moving the channel nodes into /reserved-memory is not required. > + > +As an example: > + > +/ { > + #address-cells = <2>; > + #size-cells = <2>; > + > + scmi_shm_0: sram@47ff0000 { > + compatible = "arm,scmi-shmem"; > + reg = <0x0 0x47ff0000 0x0 0x1000>; > + }; > + > + firmware { > + scmi: scmi { > + compatible = "arm,scmi-smc"; > + arm,smc-id = <0x82000002>; > + #address-cells = <1>; > + #size-cells = <0>; > + shmem = <&scmi_shm_0>; > + }; > + }; > + > + chosen { > + #address-cells = <2>; > + #size-cells = <2>; > + ranges; > + > + xen { > + #address-cells = <2>; > + #size-cells = <2>; > + ranges; > + > + xen_scmi_config { > + compatible = "xen,sci"; > + #address-cells = <2>; > + #size-cells = <2>; > + ranges; > + xen,dom0-sci-agent-id = <0>; /* dom0 agent id */ > + > + /* Xen management channel */ > + scmi_shm_1: sram@47ff1000 { > + compatible = "arm,scmi-shmem"; > + reg = <0x0 0x47ff1000 0x0 0x1000>; > + }; > + > + scmi_shm_2: sram@47ff2000 { > + compatible = "arm,scmi-shmem"; > + reg = <0x0 0x47ff2000 0x0 0x1000>; > + }; > + > + scmi_shm_3: sram@47ff3000 { > + compatible = "arm,scmi-shmem"; > + reg = <0x0 0x47ff3000 0x0 0x1000>; > + }; > + > + scmi_xen: scmi { > + compatible = "arm,scmi-smc"; /* Xen management agent func_id */ > + arm,smc-id = <0x82000003>; > + #address-cells = <1>; > + #size-cells = <0>; > + #access-controller-cells = <1>; > + shmem = <&scmi_shm_1>; /* Xen management agent shmem */ > + }; > + > + agents { > + #address-cells = <1>; > + #size-cells = <0>; > + > + agent@0 { > + compatible = "xen,scmi-agent"; > + reg = <0>; > + arm,smc-id = <0x82000002>; > + shmem = <&scmi_shm_0>; > + }; > + > + agent@2 { > + compatible = "xen,scmi-agent"; > + reg = <2>; > + arm,smc-id = <0x82000004>; > + shmem = <&scmi_shm_2>; > + }; > + > + agent@3 { > + compatible = "xen,scmi-agent"; > + reg = <3>; > + arm,smc-id = <0x82000005>; > + shmem = <&scmi_shm_3>; > + }; > + }; > + }; > + }; > + }; > +}; > + > +This keeps the host OSPM channel (SMC 0x82000002, shmem 0x47ff0000) separate > +from Xen's privileged management channel (SMC 0x82000003, shmem 0x47ff1000). > +The xen,sci subtree is omitted from Dom0's device tree. > + > +To discover a secondary agent ID, omit reg from its node. For example, > +replace only agent@2 above with: > + > +agent-domu { > + compatible = "xen,scmi-agent"; > + arm,smc-id = <0x82000004>; > + shmem = <&scmi_shm_2>; > +}; > + > +BASE_DISCOVER_AGENT is required on secondary channels only when reg is > +omitted. Xen always uses it to discover its management agent ID. Providing > +reg avoids mapping and probing the secondary channel during initialization. > +The scmi-secondary-agents and #scmi-secondary-agents-cells properties used > +by earlier versions are replaced by the agents child nodes. > + > +Dom0less example (multi-agent) > +------------------------------- > + > +Use the same Xen SCMI configuration as above. Remove xen,dom0-sci-agent-id > +if Dom0 does not use SCMI. Add xen,sci_type and xen,sci-agent-id to the > +predefined guest's node under /chosen: > + > +chosen { > + xen,domain@1 { > + compatible = "xen,domain"; > + xen,sci_type = "scmi_smc_multiagent"; > + xen,sci-agent-id = <2>; > + /* Additional domain properties (memory, cpus, kernels, etc.) */ > + }; > +}; > + > +The guest's partial DT and shared-memory mapping must use the SMC function > +ID and shared-memory channel associated with agent 2 in the agents node. > diff --git a/xen/arch/arm/dom0less-build.c b/xen/arch/arm/dom0less-build.c > index 7bbb2eafb6..01c2a5ce52 100644 > --- a/xen/arch/arm/dom0less-build.c > +++ b/xen/arch/arm/dom0less-build.c > @@ -293,6 +293,23 @@ static int __init domu_dt_sci_parse(struct dt_device_node *node, > > d_cfg->arch.arm_sci_type = XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC; > } > + else if ( !strcmp(sci_type, "scmi_smc_multiagent") ) > + { > + uint32_t agent_id = 0; > + > + if ( !IS_ENABLED(CONFIG_SCMI_SMC_MA) ) > + { > + printk(XENLOG_ERR "xen,sci_type=scmi_smc_multiagent requested, but CONFIG_SCMI_SMC_MA not set\n"); > + return -EINVAL; > + } > + > + if ( !dt_property_read_u32(node, "xen,sci-agent-id", &agent_id) || > + agent_id >= UINT8_MAX ) > + return -EINVAL; > + > + d_cfg->arch.arm_sci_type = XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC_MA; > + d_cfg->arch.arm_sci_agent_id = agent_id; > + } > else > { > printk(XENLOG_ERR "xen,sci_type in not valid (%s) for domain %s\n", > diff --git a/xen/arch/arm/domain_build.c b/xen/arch/arm/domain_build.c > index 72f6f7539a..5d42cdf6af 100644 > --- a/xen/arch/arm/domain_build.c > +++ b/xen/arch/arm/domain_build.c > @@ -88,6 +88,39 @@ int __init parse_arch_dom0_param(const char *s, const char *e) > return -EINVAL; > } > > +#ifdef CONFIG_SCMI_SMC_MA > +/* SCMI agent ID for dom0 obtained from xen,sci container */ > +#define SCMI_AGENT_ID_INVALID UINT8_MAX > + > +static uint8_t __init get_dom0_scmi_agent_id(void) > +{ > + const struct dt_device_node *config_node; > + u32 val; > + const struct dt_property *prop; > + > + config_node = dt_find_compatible_node(NULL, NULL, "xen,sci"); > + if ( !config_node ) > + return SCMI_AGENT_ID_INVALID; > + > + prop = dt_find_property(config_node, "xen,dom0-sci-agent-id", NULL); > + if ( !prop ) > + return SCMI_AGENT_ID_INVALID; > + > + if ( !dt_property_read_u32(config_node, "xen,dom0-sci-agent-id", &val) ) > + return SCMI_AGENT_ID_INVALID; > + > + if ( val >= SCMI_AGENT_ID_INVALID ) > + { > + printk(XENLOG_WARNING > + "Invalid xen,dom0-sci-agent-id=%u, SCMI disabled for Dom0\n", > + val); > + return SCMI_AGENT_ID_INVALID; > + } > + > + return val; > +} > +#endif /* CONFIG_SCMI_SMC_MA */ > + > /* Override macros from asm/page.h to make them work with mfn_t */ > #undef virt_to_mfn > #define virt_to_mfn(va) _mfn(__virt_to_mfn(va)) > @@ -1480,6 +1513,7 @@ static int __init handle_node(struct domain *d, struct kernel_info *kinfo, > DT_MATCH_TYPE("memory"), > /* The memory mapped timer is not supported by Xen. */ > DT_MATCH_COMPATIBLE("arm,armv7-timer-mem"), > + DT_MATCH_COMPATIBLE("xen,sci"), > { /* sentinel */ }, > }; > static const struct dt_device_match timer_matches[] __initconst = > @@ -1975,6 +2009,15 @@ void __init create_dom0(void) > dom0_cfg.arch.tee_type = tee_get_type(); > dom0_cfg.max_vcpus = dom0_max_vcpus(); > > +#ifdef CONFIG_SCMI_SMC_MA > + /* Set up SCMI agent ID if provided in the xen,sci container */ > + dom0_cfg.arch.arm_sci_agent_id = get_dom0_scmi_agent_id(); > + dom0_cfg.arch.arm_sci_type = (dom0_cfg.arch.arm_sci_agent_id != > + SCMI_AGENT_ID_INVALID) ? > + XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC_MA : > + XEN_DOMCTL_CONFIG_ARM_SCI_NONE; > +#endif > + > if ( iommu_enabled ) > dom0_cfg.flags |= XEN_DOMCTL_CDF_iommu; > > diff --git a/xen/arch/arm/firmware/Kconfig b/xen/arch/arm/firmware/Kconfig > index 5c5f0880c4..972cd9b173 100644 > --- a/xen/arch/arm/firmware/Kconfig > +++ b/xen/arch/arm/firmware/Kconfig > @@ -29,6 +29,18 @@ config SCMI_SMC > driver domain. > Use with EL3 firmware which supports only single SCMI OSPM agent. > > +config SCMI_SMC_MA > + bool "Enable ARM SCMI SMC multi-agent driver" > + depends on ARM_64 > + select ARM_SCI > + help > + Enables SCMI SMC/HVC multi-agent in XEN to pass SCMI requests from Domains > + to EL3 firmware (TF-A) which supports multi-agent feature. > + This feature allows to enable SCMI per Domain using unique SCMI agent_id, > + so Domain is identified by EL3 firmware as an SCMI Agent and can access > + allowed platform resources through dedicated SMC/HVC Shared memory based > + transport. > + > endchoice > > endmenu > diff --git a/xen/arch/arm/firmware/Makefile b/xen/arch/arm/firmware/Makefile > index 71bdefc24a..37927e690e 100644 > --- a/xen/arch/arm/firmware/Makefile > +++ b/xen/arch/arm/firmware/Makefile > @@ -1,2 +1,3 @@ > obj-$(CONFIG_ARM_SCI) += sci.o > obj-$(CONFIG_SCMI_SMC) += scmi-smc.o > +obj-$(CONFIG_SCMI_SMC_MA) += scmi-shmem.o scmi-smc-multiagent.o > diff --git a/xen/arch/arm/firmware/scmi-proto.h b/xen/arch/arm/firmware/scmi-proto.h > new file mode 100644 > index 0000000000..49f63cfc0a > --- /dev/null > +++ b/xen/arch/arm/firmware/scmi-proto.h > @@ -0,0 +1,164 @@ > +/* SPDX-License-Identifier: GPL-2.0-only */ > +/* > + * Arm System Control and Management Interface definitions > + * Version 3.0 (DEN0056C) > + * > + * Copyright (c) 2025 EPAM Systems > + */ > + > +#ifndef ARM_FIRMWARE_SCMI_PROTO_H_ > +#define ARM_FIRMWARE_SCMI_PROTO_H_ > + > +#include <xen/stdint.h> > + > +#define SCMI_SHORT_NAME_MAX_SIZE 16 > + > +/* SCMI status codes. See section 4.1.4 */ > +#define SCMI_SUCCESS 0 > +#define SCMI_NOT_SUPPORTED (-1) > +#define SCMI_INVALID_PARAMETERS (-2) > +#define SCMI_DENIED (-3) > +#define SCMI_NOT_FOUND (-4) > +#define SCMI_OUT_OF_RANGE (-5) > +#define SCMI_BUSY (-6) > +#define SCMI_COMMS_ERROR (-7) > +#define SCMI_GENERIC_ERROR (-8) > +#define SCMI_HARDWARE_ERROR (-9) > +#define SCMI_PROTOCOL_ERROR (-10) > + > +/* Protocol IDs */ > +#define SCMI_BASE_PROTOCOL 0x10 > + > +/* Base protocol message IDs */ > +#define SCMI_BASE_PROTOCOL_VERSION 0x0 > +#define SCMI_BASE_PROTOCOL_ATTIBUTES 0x1 > +#define SCMI_BASE_PROTOCOL_MESSAGE_ATTRIBUTES 0x2 > +#define SCMI_BASE_DISCOVER_AGENT 0x7 > +#define SCMI_BASE_SET_DEVICE_PERMISSIONS 0x9 > +#define SCMI_BASE_RESET_AGENT_CONFIGURATION 0xB > + > +typedef struct scmi_msg_header { > + uint8_t id; > + uint8_t type; > + uint8_t protocol; > + uint32_t status; > +} scmi_msg_header_t; > + > +/* Table 2 Message header format */ > +#define SCMI_HDR_ID GENMASK(7, 0) > +#define SCMI_HDR_TYPE GENMASK(9, 8) > +#define SCMI_HDR_PROTO GENMASK(17, 10) > + > +#define SCMI_FIELD_GET(_mask, _reg) \ > + ((typeof(_mask))(((_reg) & (_mask)) >> (ffs64(_mask) - 1))) > +#define SCMI_FIELD_PREP(_mask, _val) \ > + (((typeof(_mask))(_val) << (ffs64(_mask) - 1)) & (_mask)) > + > +static inline uint32_t pack_scmi_header(scmi_msg_header_t *hdr) > +{ > + return SCMI_FIELD_PREP(SCMI_HDR_ID, hdr->id) | > + SCMI_FIELD_PREP(SCMI_HDR_TYPE, hdr->type) | > + SCMI_FIELD_PREP(SCMI_HDR_PROTO, hdr->protocol); > +} > + > +static inline void unpack_scmi_header(uint32_t msg_hdr, scmi_msg_header_t *hdr) > +{ > + hdr->id = SCMI_FIELD_GET(SCMI_HDR_ID, msg_hdr); > + hdr->type = SCMI_FIELD_GET(SCMI_HDR_TYPE, msg_hdr); > + hdr->protocol = SCMI_FIELD_GET(SCMI_HDR_PROTO, msg_hdr); > +} > + > +static inline int scmi_to_xen_errno(int scmi_status) > +{ > + if ( scmi_status == SCMI_SUCCESS ) > + return 0; > + > + switch ( scmi_status ) > + { > + case SCMI_NOT_SUPPORTED: > + return -EOPNOTSUPP; > + case SCMI_INVALID_PARAMETERS: > + return -EINVAL; > + case SCMI_DENIED: > + return -EACCES; > + case SCMI_NOT_FOUND: > + return -ENOENT; > + case SCMI_OUT_OF_RANGE: > + return -ERANGE; > + case SCMI_BUSY: > + return -EBUSY; > + case SCMI_COMMS_ERROR: > + return -ENOTCONN; > + case SCMI_GENERIC_ERROR: > + return -EIO; > + case SCMI_HARDWARE_ERROR: > + return -ENXIO; > + case SCMI_PROTOCOL_ERROR: > + return -EBADMSG; > + default: > + return -EINVAL; > + } > +} > + > +/* PROTOCOL_VERSION */ > +#define SCMI_VERSION_MINOR GENMASK(15, 0) > +#define SCMI_VERSION_MAJOR GENMASK(31, 16) > + > +struct scmi_msg_prot_version_p2a { > + uint32_t version; > +} __packed; > + > +/* BASE PROTOCOL_ATTRIBUTES */ > +#define SCMI_BASE_ATTR_NUM_PROTO GENMASK(7, 0) > +#define SCMI_BASE_ATTR_NUM_AGENT GENMASK(15, 8) > + > +struct scmi_msg_base_attributes_p2a { > + uint32_t attributes; > +} __packed; > + > +/* > + * BASE_DISCOVER_AGENT > + */ > +#define SCMI_BASE_AGENT_ID_OWN 0xFFFFFFFF > + > +struct scmi_msg_base_discover_agent_a2p { > + uint32_t agent_id; > +} __packed; > + > +struct scmi_msg_base_discover_agent_p2a { > + uint32_t agent_id; > + char name[SCMI_SHORT_NAME_MAX_SIZE]; > +} __packed; > + > +/* > + * BASE_SET_DEVICE_PERMISSIONS > + */ > +#define SCMI_BASE_DEVICE_ACCESS_ALLOW BIT(0, UL) > + > +struct scmi_msg_base_set_device_permissions_a2p { > + uint32_t agent_id; > + uint32_t device_id; > + uint32_t flags; > +} __packed; > + > +/* > + * BASE_RESET_AGENT_CONFIGURATION > + */ > +#define SCMI_BASE_AGENT_PERMISSIONS_RESET BIT(0, UL) > + > +struct scmi_msg_base_reset_agent_cfg_a2p { > + uint32_t agent_id; > + uint32_t flags; > +} __packed; > + > +#endif /* ARM_FIRMWARE_SCMI_PROTO_H_ */ > + > +/* > + * Local variables: > + * mode: C > + * c-file-style: "BSD" > + * c-basic-offset: 4 > + * tab-width: 4 > + * indent-tabs-mode: nil > + * End: > + */ > diff --git a/xen/arch/arm/firmware/scmi-shmem.c b/xen/arch/arm/firmware/scmi-shmem.c > new file mode 100644 > index 0000000000..e36745a85e > --- /dev/null > +++ b/xen/arch/arm/firmware/scmi-shmem.c > @@ -0,0 +1,118 @@ > +/* SPDX-License-Identifier: GPL-2.0-only */ > +/* > + * SMC/HVC shmem transport implementation used by > + * SCI SCMI multi-agent driver. > + * > + * Oleksii Moisieiev <oleksii_moisieiev@epam.com> > + * Copyright (c) 2025 EPAM Systems > + */ > + > +#include <xen/err.h> > +#include <xen/io.h> > +#include <asm/io.h> > + > +#include "scmi-proto.h" > +#include "scmi-shmem.h" > + > +static inline int > +shmem_channel_status(const volatile struct scmi_shared_mem __iomem *shmem) > +{ > + return (readl(&shmem->channel_status) & > + SCMI_SHMEM_CHAN_STAT_CHANNEL_FREE) ? 0 : -EBUSY; > +} > + > +int shmem_put_message(volatile struct scmi_shared_mem __iomem *shmem, > + scmi_msg_header_t *hdr, void *data, unsigned int len) > +{ > + int ret; > + > + if ( (len + offsetof(struct scmi_shared_mem, msg_payload)) > > + SCMI_SHMEM_MAPPED_SIZE ) > + { > + printk(XENLOG_ERR "scmi: Wrong size of smc message. Data is invalid\n"); "of the SMC message". Also, looks like it is over 80 characters. > + return -EINVAL; > + } > + > + ret = shmem_channel_status(shmem); > + if ( ret ) > + return ret; > + > + writel_relaxed(0x0, &shmem->channel_status); > + /* Writing 0x0 right now, but "shmem"_FLAG_INTR_ENABLED can be set */ > + writel_relaxed(0x0, &shmem->flags); > + writel_relaxed(sizeof(shmem->msg_header) + len, &shmem->length); > + writel(pack_scmi_header(hdr), &shmem->msg_header); Why do you mix writel() and writel_relaxed() here? > + > + if ( len > 0 && data ) > + memcpy_toio(shmem->msg_payload, data, len); > + > + return 0; > +} > + > +int shmem_get_response(const volatile struct scmi_shared_mem __iomem *shmem, > + scmi_msg_header_t *hdr, void *data, unsigned int len) > +{ > + int recv_len; > + int ret; > + /* > + * First word of msg_payload carries the returned status; exclude it from > + * recv_len so only the protocol payload is copied back to the caller. > + */ > + int pad = sizeof(hdr->status); > + > + if ( len >= SCMI_SHMEM_MAPPED_SIZE - > + offsetof(struct scmi_shared_mem, msg_payload) ) > + { > + printk(XENLOG_ERR > + "scmi: Wrong size of input smc message. Data may be invalid\n"); "of the input SMC message" > + return -EINVAL; > + } > + > + ret = shmem_channel_status(shmem); > + if ( ret ) > + return ret; > + > + recv_len = readl(&shmem->length) - sizeof(shmem->msg_header); > + > + if ( recv_len < 0 ) > + { > + printk(XENLOG_ERR > + "scmi: Wrong size of smc message. Data may be invalid\n"); "of the SMC message" Also, I am wondering if you really need to print this error messages in the first place. > + return -EINVAL; > + } > + > + unpack_scmi_header(readl(&shmem->msg_header), hdr); > + > + hdr->status = readl(&shmem->msg_payload); > + recv_len = recv_len > pad ? recv_len - pad : 0; > + > + ret = scmi_to_xen_errno(hdr->status); > + if ( ret ) > + { > + printk(XENLOG_DEBUG "scmi: Error received: %d\n", ret); > + return ret; > + } > + > + if ( recv_len > len ) > + { > + printk(XENLOG_ERR > + "scmi: Not enough buffer for message %d, expecting %d\n", > + recv_len, len); > + return -EINVAL; > + } > + > + if ( recv_len > 0 ) > + memcpy_fromio(data, shmem->msg_payload + pad, recv_len); > + > + return 0; > +} > + > +/* > + * Local variables: > + * mode: C > + * c-file-style: "BSD" > + * c-basic-offset: 4 > + * tab-width: 4 > + * indent-tabs-mode: nil > + * End: > + */ > diff --git a/xen/arch/arm/firmware/scmi-shmem.h b/xen/arch/arm/firmware/scmi-shmem.h > new file mode 100644 > index 0000000000..722263aa77 > --- /dev/null > +++ b/xen/arch/arm/firmware/scmi-shmem.h > @@ -0,0 +1,45 @@ > +/* SPDX-License-Identifier: GPL-2.0-only */ > +/* > + * Arm System Control and Management Interface definitions > + * Version 3.0 (DEN0056C) > + * Shared Memory based Transport > + * > + * Copyright (c) 2024 EPAM Systems > + */ > + > +#ifndef ARM_FIRMWARE_SCMI_SHMEM_H_ > +#define ARM_FIRMWARE_SCMI_SHMEM_H_ > + > +#include <xen/stdint.h> > + > +#define SCMI_SHMEM_CHAN_STAT_CHANNEL_FREE BIT(0, UL) > +#define SCMI_SHMEM_CHAN_STAT_CHANNEL_ERROR BIT(1, UL) > + > +struct scmi_shared_mem { > + uint32_t reserved; > + uint32_t channel_status; > + uint32_t reserved1[2]; > + uint32_t flags; > + uint32_t length; > + uint32_t msg_header; > + uint8_t msg_payload[]; > +}; Should it be __packed? > + > +#define SCMI_SHMEM_MAPPED_SIZE PAGE_SIZE > + > +int shmem_put_message(volatile struct scmi_shared_mem __iomem *shmem, I am not sure if volatile makes sense here, taking into account that this is __iomem anyway. Also, shmem_put_message sounds to generic... Maybe scmi_put_message? > + scmi_msg_header_t *hdr, void *data, unsigned int len); > + > +int shmem_get_response(const volatile struct scmi_shared_mem __iomem *shmem, > + scmi_msg_header_t *hdr, void *data, unsigned int len); Basically the same comments as for previous one. > +#endif /* ARM_FIRMWARE_SCMI_SHMEM_H_ */ > + > +/* > + * Local variables: > + * mode: C > + * c-file-style: "BSD" > + * c-basic-offset: 4 > + * tab-width: 4 > + * indent-tabs-mode: nil > + * End: > + */ > diff --git a/xen/arch/arm/firmware/scmi-smc-multiagent.c b/xen/arch/arm/firmware/scmi-smc-multiagent.c > new file mode 100644 > index 0000000000..6c46f6cd36 > --- /dev/null > +++ b/xen/arch/arm/firmware/scmi-smc-multiagent.c > @@ -0,0 +1,792 @@ > +/* SPDX-License-Identifier: GPL-2.0-only */ > +/* > + * SCI SCMI multi-agent driver, using SMC/HVC shmem as transport. > + * > + * Oleksii Moisieiev <oleksii_moisieiev@epam.com> > + * Copyright (c) 2025 EPAM Systems > + */ > + > +#include <xen/acpi.h> > + > +#include <xen/device_tree.h> > +#include <xen/init.h> > +#include <xen/iocap.h> > +#include <xen/err.h> > +#include <xen/libfdt/libfdt.h> > +#include <xen/string.h> > +#include <xen/param.h> > +#include <xen/sched.h> > +#include <xen/vmap.h> > + > +#include <asm/firmware/sci.h> > +#include <asm/smccc.h> > + > +#include "scmi-proto.h" > +#include "scmi-shmem.h" > + > +struct scmi_channel { > + uint32_t agent_id; If all agent_ids large than UINT8_MAX are invalid, why do you need uint32_t here? > + uint32_t func_id; > + domid_t domain_id; > + uint64_t paddr; paddr_t > + struct scmi_shared_mem __iomem *shmem; > + spinlock_t lock; > + struct list_head list; > +}; > + > +struct scmi_data { > + struct list_head channel_list; > + spinlock_t channel_list_lock; > + uint32_t func_id; > + bool initialized; > + uint32_t shmem_phandle; > + uint32_t hyp_channel_agent_id; > + struct dt_device_node *dt_dev; > +}; > + > +static struct scmi_data scmi_data; > + > +static bool scmi_is_under_xen_sci(const struct dt_device_node *node) > +{ > + const struct dt_device_node *p; > + > + for ( p = node->parent; p; p = p->parent ) > + if ( dt_device_is_compatible(p, "xen,sci") ) > + return true; > + > + return false; > +} > + > +static int send_smc_message(struct scmi_channel *chan_info, > + scmi_msg_header_t *hdr, void *data, int len) > +{ > + struct arm_smccc_res resp; > + int ret; > + > + ret = shmem_put_message(chan_info->shmem, hdr, data, len); > + if ( ret ) > + return ret; > + > + resp = arm_smccc_1_1_smc(chan_info->func_id, 0, 0, 0, 0, 0, 0, 0); > + > + if ( resp.a0 == (unsigned long)ARM_SMCCC_INVALID_PARAMETER ) > + return -EINVAL; > + > + if ( resp.a0 ) > + return -EOPNOTSUPP; > + > + return 0; > +} > + > +static int do_smc_xfer(struct scmi_channel *chan_info, scmi_msg_header_t *hdr, > + void *tx_data, int tx_size, void *rx_data, int rx_size) > +{ > + int ret = 0; > + > + ASSERT(chan_info && chan_info->shmem); > + > + if ( !hdr ) > + return -EINVAL; > + > + spin_lock(&chan_info->lock); > + > + printk(XENLOG_DEBUG > + "scmi: agent_id = %d msg_id = %x type = %d, proto = %x\n", > + chan_info->agent_id, hdr->id, hdr->type, hdr->protocol); > + > + ret = send_smc_message(chan_info, hdr, tx_data, tx_size); > + if ( ret ) > + goto clean; > + > + ret = shmem_get_response(chan_info->shmem, hdr, rx_data, rx_size); > + > +clean: > + printk(XENLOG_DEBUG > + "scmi: get smc response agent_id = %d msg_id = %x proto = %x res=%d\n", > + chan_info->agent_id, hdr->id, hdr->protocol, ret); You will print this even if send_smc_message() fails. I think it should go before "clean:" label. Also, maybe "cleanup:" ? > + > + spin_unlock(&chan_info->lock); > + > + return ret; > +} > + > +static struct scmi_channel *get_channel_by_id(uint32_t agent_id) > +{ > + struct scmi_channel *curr; > + bool found = false; > + > + spin_lock(&scmi_data.channel_list_lock); > + list_for_each_entry(curr, &scmi_data.channel_list, list) > + { > + if ( curr->agent_id == agent_id ) > + { > + found = true; > + break; > + } > + } > + > + spin_unlock(&scmi_data.channel_list_lock); > + if ( found ) > + return curr; > + > + return NULL; > +} > + > +static struct scmi_channel *acquire_scmi_channel(struct domain *d, > + uint32_t agent_id) > +{ > + struct scmi_channel *curr; > + struct scmi_channel *ret = ERR_PTR(-ENOENT); > + > + spin_lock(&scmi_data.channel_list_lock); > + list_for_each_entry(curr, &scmi_data.channel_list, list) > + { > + if ( curr->agent_id == agent_id ) > + { > + if ( curr->domain_id != DOMID_INVALID ) > + { > + ret = ERR_PTR(-EEXIST); > + break; > + } > + > + curr->domain_id = d->domain_id; > + ret = curr; > + break; > + } > + } > + > + spin_unlock(&scmi_data.channel_list_lock); > + > + return ret; > +} > + > +static void relinquish_scmi_channel(struct scmi_channel *channel) > +{ > + ASSERT(channel != NULL); > + > + spin_lock(&scmi_data.channel_list_lock); > + channel->domain_id = DOMID_INVALID; > + spin_unlock(&scmi_data.channel_list_lock); > +} > + > +static int map_channel_memory(struct scmi_channel *channel) > +{ > + ASSERT(channel && channel->paddr); > + channel->shmem = ioremap_nocache(channel->paddr, SCMI_SHMEM_MAPPED_SIZE); > + if ( !channel->shmem ) > + return -ENOMEM; > + > + channel->shmem->channel_status = SCMI_SHMEM_CHAN_STAT_CHANNEL_FREE; > + printk(XENLOG_DEBUG "scmi: Got shmem %lx after vmap %p\n", channel->paddr, You will need proper format, not "%lx" for channel->paddr > + channel->shmem); > + > + return 0; > +} > + > +static void unmap_channel_memory(struct scmi_channel *channel) > +{ > + ASSERT(channel); > + > + if ( !channel->shmem ) > + return; > + > + iounmap(channel->shmem); > + channel->shmem = NULL; > +} > + > +static struct scmi_channel *smc_create_channel(uint32_t agent_id, > + uint32_t func_id, uint64_t addr) > +{ > + struct scmi_channel *channel, *curr; > + > + spin_lock(&scmi_data.channel_list_lock); > + > + /* Check if channel already exists while holding the lock */ > + list_for_each_entry(curr, &scmi_data.channel_list, list) > + { > + if ( curr->agent_id == agent_id ) > + { > + spin_unlock(&scmi_data.channel_list_lock); > + return ERR_PTR(-EEXIST); > + } > + } > + > + channel = xmalloc(struct scmi_channel); > + if ( !channel ) > + { > + spin_unlock(&scmi_data.channel_list_lock); > + return ERR_PTR(-ENOMEM); > + } > + > + spin_lock_init(&channel->lock); > + channel->agent_id = agent_id; > + channel->func_id = func_id; > + channel->domain_id = DOMID_INVALID; > + channel->shmem = NULL; > + channel->paddr = addr; > + list_add_tail(&channel->list, &scmi_data.channel_list); > + > + spin_unlock(&scmi_data.channel_list_lock); > + return channel; > +} > + > +static void free_channel_list(void) > +{ > + struct scmi_channel *curr, *_curr; > + /* > + * Called only on the __init error path, before any runtime users exist, Then you can put this function into __init as well > + * so no other thread is iterating channel_list. Keep the lock for the > + * whole drain for clarity and to avoid misleading readers about possible > + * concurrent access. > + */ > + spin_lock(&scmi_data.channel_list_lock); > + list_for_each_entry_safe(curr, _curr, &scmi_data.channel_list, list) > + { > + list_del(&curr->list); > + xfree(curr); > + } > + spin_unlock(&scmi_data.channel_list_lock); > +} > + > +static int __init > +scmi_dt_read_channel_addr(struct dt_device_node *scmi_node, u64 *addr, > + u64 *size) > +{ > + struct dt_device_node *shmem_node; > + uint32_t phandle; > + int ret; > + > + ret = dt_property_read_variable_u32_array(scmi_node, "shmem", &phandle, > + 1, 1); > + if ( ret < 0 ) > + return ret; > + > + shmem_node = dt_find_node_by_phandle(phandle); > + if ( !shmem_node || > + !dt_device_is_compatible(shmem_node, "arm,scmi-shmem") ) > + return -EINVAL; > + > + ret = dt_device_get_address(shmem_node, 0, addr, size); > + if ( ret ) > + return ret; > + > + if ( !*size || !IS_ALIGNED(*size, SCMI_SHMEM_MAPPED_SIZE) || > + !IS_ALIGNED(*addr, SCMI_SHMEM_MAPPED_SIZE) ) > + { > + printk(XENLOG_ERR "scmi: invalid shmem range for %s\n", > + scmi_node->full_name); > + return -EINVAL; > + } > + > + return 0; > +} > + > +/* > + * Handle Dom0 SCMI specific DT nodes > + * > + * Make a decision on copying SCMI specific nodes into Dom0 device tree. > + * For SCMI multi-agent case: > + * - shmem nodes will not be copied and generated instead if SCMI > + * is enabled for Dom0 > + * - scmi node will be copied if SCMI is enabled for Dom0 > + */ > +static bool scmi_dt_handle_node(struct domain *d, struct dt_device_node *node) Also __init? > +{ > + static const struct dt_device_match shmem_matches[] __initconst = { > + DT_MATCH_COMPATIBLE("arm,scmi-shmem"), > + { /* sentinel */ }, > + }; > + static const struct dt_device_match scmi_matches[] __initconst = { > + DT_MATCH_PATH("/firmware/scmi"), > + { /* sentinel */ }, > + }; > + > + if ( !scmi_data.initialized ) > + return false; > + > + /* skip scmi shmem node for dom0 if scmi not enabled */ > + if ( dt_match_node(shmem_matches, node) && !sci_domain_is_enabled(d) ) > + { > + dt_dprintk(" Skip scmi shmem node\n"); > + return true; > + } > + > + /* drop scmi if not enabled */ > + if ( dt_match_node(scmi_matches, node) && !sci_domain_is_enabled(d) ) > + { > + dt_dprintk(" Skip scmi node\n"); > + return true; > + } > + > + return false; > +} > + > +static int scmi_assign_device(uint32_t agent_id, uint32_t device_id, > + uint32_t flags) > +{ > + struct scmi_msg_base_set_device_permissions_a2p tx; > + struct scmi_channel *channel; > + scmi_msg_header_t hdr; > + > + channel = get_channel_by_id(scmi_data.hyp_channel_agent_id); > + if ( !channel ) > + return -EINVAL; > + > + hdr.id = SCMI_BASE_SET_DEVICE_PERMISSIONS; > + hdr.type = 0; > + hdr.protocol = SCMI_BASE_PROTOCOL; > + > + tx.agent_id = agent_id; > + tx.device_id = device_id; > + tx.flags = flags; > + > + return do_smc_xfer(channel, &hdr, &tx, sizeof(tx), NULL, 0); > +} > + > +static int scmi_dt_assign_device(struct domain *d, > + struct dt_phandle_args *ac_spec) > +{ > + struct scmi_channel *agent_channel; > + uint32_t scmi_device_id = ac_spec->args[0]; > + int ret; > + > + if ( !d->arch.sci_data ) > + return 0; > + > + /* The access-controllers is specified for DT dev, but it's not a SCMI */ > + if ( !scmi_data.dt_dev || > + !dt_node_path_is_equal(ac_spec->np, scmi_data.dt_dev->full_name) ) > + return 0; > + > + agent_channel = d->arch.sci_data; > + > + spin_lock(&agent_channel->lock); > + > + ret = scmi_assign_device(agent_channel->agent_id, scmi_device_id, > + SCMI_BASE_DEVICE_ACCESS_ALLOW); > + if ( ret ) > + { > + printk(XENLOG_ERR > + "scmi: could not assign dev for %pd agent:%d dev_id:%u (%d)", > + d, agent_channel->agent_id, scmi_device_id, ret); > + } > + > + spin_unlock(&agent_channel->lock); > + return ret; > +} > + > +static int collect_agent_id(struct scmi_channel *agent_channel) > +{ > + int ret; > + scmi_msg_header_t hdr; > + struct scmi_msg_base_discover_agent_p2a da_rx; > + struct scmi_msg_base_discover_agent_a2p da_tx; > + > + ret = map_channel_memory(agent_channel); > + if ( ret ) > + return ret; > + > + hdr.id = SCMI_BASE_DISCOVER_AGENT; > + hdr.type = 0; > + hdr.protocol = SCMI_BASE_PROTOCOL; > + > + da_tx.agent_id = agent_channel->agent_id; > + > + ret = do_smc_xfer(agent_channel, &hdr, &da_tx, sizeof(da_tx), &da_rx, > + sizeof(da_rx)); > + if ( agent_channel->domain_id != DOMID_XEN ) > + unmap_channel_memory(agent_channel); > + if ( ret ) On error, Xen's channel will be mapped still. Is this intended behaviour? > + return ret; > + > + printk(XENLOG_DEBUG "id=0x%x name=%s\n", da_rx.agent_id, da_rx.name); > + agent_channel->agent_id = da_rx.agent_id; > + return 0; > +} > + > +static int __init collect_agents(struct dt_device_node *scmi_node) > +{ > + struct dt_device_node *agents, *node; > + unsigned int count = 0; > + > + /* Use the configuration containing this management channel. */ > + for ( agents = scmi_node->parent; agents; agents = agents->parent ) > + if ( dt_device_is_compatible(agents, "xen,sci") ) > + break; > + > + ASSERT(agents); > + > + dt_for_each_child_node(agents, node) > + if ( dt_node_name_is_equal(node, "agents") ) > + break; > + > + if ( !node ) > + { > + printk(XENLOG_ERR "scmi: missing agents node\n"); > + return -EINVAL; > + } > + agents = node; > + > + if ( dt_child_n_addr_cells(agents) != 1 || > + dt_child_n_size_cells(agents) != 0 ) > + return -EINVAL; > + > + dt_for_each_child_node(agents, node) > + { > + struct scmi_channel agent = { > + .agent_id = SCMI_BASE_AGENT_ID_OWN, > + .domain_id = DOMID_INVALID, > + }; > + struct scmi_channel *channel; > + bool discover; > + uint32_t len; > + u64 size; uint64_t > + int ret; > + > + if ( !dt_device_is_compatible(node, "xen,scmi-agent") || > + !dt_device_is_available(node) ) > + continue; > + > + discover = !dt_find_property(node, "reg", &len); > + if ( !discover ) > + { > + if ( len != sizeof(uint32_t) || > + !dt_property_read_u32(node, "reg", &agent.agent_id) ) > + return -EINVAL; > + } > + > + if ( !dt_property_read_u32(node, "arm,smc-id", &agent.func_id) ) > + return -EINVAL; > + > + ret = scmi_dt_read_channel_addr(node, &agent.paddr, &size); > + if ( ret ) > + return ret; > + > + if ( discover ) > + { > + /* Resolve the ID before inserting it into the channel list. */ > + spin_lock_init(&agent.lock); > + ret = collect_agent_id(&agent); > + if ( ret ) > + return ret; > + } > + > + /* Guest configurations encode agent IDs in an eight-bit field. */ > + if ( agent.agent_id >= UINT8_MAX ) > + return -EINVAL; > + > + channel = smc_create_channel(agent.agent_id, agent.func_id, > + agent.paddr); > + if ( IS_ERR(channel) ) > + { > + printk(XENLOG_ERR "scmi: Could not create channel for %s: %ld\n", > + node->full_name, PTR_ERR(channel)); > + return PTR_ERR(channel); > + } > + > + count++; > + printk(XENLOG_DEBUG "scmi: Agent %u SMC %X addr %lx\n", > + agent.agent_id, agent.func_id, (unsigned long)agent.paddr); Why cast to unsigned long? > + } > + > + return count ? 0 : -EINVAL; > +} > + > +static int scmi_domain_init(struct domain *d, > + struct xen_domctl_createdomain *config) > +{ > + struct scmi_channel *channel; > + int ret; > + > + if ( !scmi_data.initialized ) > + return 0; > + > + /* > + * SCMI support is configured via: > + * - For dom0: xen,dom0-sci-agent-id property under the xen,sci container > + * - For dom0less: xen,sci-agent-id in the domain node > + * The config->arch.arm_sci_type and config->arch.arm_sci_agent_id > + * are already set by domain_build.c or dom0less-build.c > + */ > + > + if ( config->arch.arm_sci_type == XEN_DOMCTL_CONFIG_ARM_SCI_NONE ) > + return 0; > + > + channel = acquire_scmi_channel(d, config->arch.arm_sci_agent_id); > + if ( IS_ERR(channel) ) > + { > + printk(XENLOG_ERR > + "scmi: Failed to acquire SCMI channel for agent_id %u: %ld\n", > + config->arch.arm_sci_agent_id, PTR_ERR(channel)); > + return PTR_ERR(channel); > + } > + > + printk(XENLOG_INFO > + "scmi: Acquire channel id = 0x%x, domain_id = %d paddr = 0x%lx\n", > + channel->agent_id, channel->domain_id, channel->paddr); > + > + /* > + * Dom0 (if present) needs to have an access to the guest memory range > + * to satisfy iomem_access_permitted() check in XEN_DOMCTL_iomem_permission > + * domctl. > + */ > + if ( hardware_domain && !is_hardware_domain(d) ) > + { > + ret = iomem_permit_access(hardware_domain, paddr_to_pfn(channel->paddr), > + paddr_to_pfn(channel->paddr + > + SCMI_SHMEM_MAPPED_SIZE - 1)); > + if ( ret ) > + goto error; > + } > + > + d->arch.sci_data = channel; > + d->arch.sci_enabled = true; > + > + return 0; > + > +error: > + relinquish_scmi_channel(channel); > + return ret; > +} > + > +int scmi_domain_sanitise_config(const struct xen_domctl_createdomain *config) > +{ > + if ( config->arch.arm_sci_type != XEN_DOMCTL_CONFIG_ARM_SCI_NONE && > + config->arch.arm_sci_type != XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC_MA ) > + { > + dprintk(XENLOG_INFO, "scmi: Unsupported ARM_SCI type\n"); > + return -EINVAL; > + } > + > + return 0; > +} > + > +static int scmi_relinquish_resources(struct domain *d) > +{ > + int ret; > + struct scmi_channel *channel, *agent_channel; > + scmi_msg_header_t hdr; > + struct scmi_msg_base_reset_agent_cfg_a2p tx; > + > + if ( !d->arch.sci_data ) > + return 0; > + > + agent_channel = d->arch.sci_data; > + > + spin_lock(&agent_channel->lock); > + tx.agent_id = agent_channel->agent_id; > + spin_unlock(&agent_channel->lock); > + > + channel = get_channel_by_id(scmi_data.hyp_channel_agent_id); > + if ( !channel ) > + { > + printk(XENLOG_ERR > + "scmi: Unable to get Hypervisor scmi channel for domain %d\n", > + d->domain_id); > + return -EINVAL; > + } > + > + hdr.id = SCMI_BASE_RESET_AGENT_CONFIGURATION; > + hdr.type = 0; > + hdr.protocol = SCMI_BASE_PROTOCOL; > + > + tx.flags = SCMI_BASE_AGENT_PERMISSIONS_RESET; > + > + ret = do_smc_xfer(channel, &hdr, &tx, sizeof(tx), NULL, 0); > + if ( ret == -EOPNOTSUPP ) > + return 0; > + > + return ret; > +} > + > +static void scmi_domain_destroy(struct domain *d) > +{ > + struct scmi_channel *channel; > + > + if ( !d->arch.sci_data ) > + return; > + > + channel = d->arch.sci_data; > + spin_lock(&channel->lock); > + > + relinquish_scmi_channel(channel); > + printk(XENLOG_DEBUG "scmi: Free domain %d\n", d->domain_id); You can use "%pd" > + > + d->arch.sci_data = NULL; > + d->arch.sci_enabled = false; > + > + spin_unlock(&channel->lock); > +} > + > +static bool scmi_handle_call(struct cpu_user_regs *regs) > +{ > + uint32_t fid = (uint32_t)get_user_reg(regs, 0); > + struct scmi_channel *agent_channel; > + struct domain *d = current->domain; > + bool res = false; > + > + if ( (!sci_domain_is_enabled(d)) || (!d->arch.sci_data) ) > + return false; > + > + agent_channel = d->arch.sci_data; > + spin_lock(&agent_channel->lock); > + > + if ( agent_channel->func_id != fid ) > + { > + res = false; > + goto unlock; > + } > + > + arm_smccc_guest_smc(regs); > + res = true; > +unlock: > + spin_unlock(&agent_channel->lock); > + > + return res; > +} > + > +static const struct sci_mediator_ops scmi_ops = { > + .domain_init = scmi_domain_init, > + .domain_destroy = scmi_domain_destroy, > + .relinquish_resources = scmi_relinquish_resources, > + .handle_call = scmi_handle_call, > + .dom0_dt_handle_node = scmi_dt_handle_node, > + .domain_sanitise_config = scmi_domain_sanitise_config, > + .assign_dt_device = scmi_dt_assign_device, > +}; > + > +static int __init scmi_check_smccc_ver(void) > +{ > + if ( smccc_ver < ARM_SMCCC_VERSION_1_1 ) > + { > + printk(XENLOG_WARNING > + "scmi: No SMCCC 1.1 support, SCMI calls forwarding disabled\n"); > + return -ENOSYS; > + } > + > + return 0; > +} > + > +static int __init scmi_dt_hyp_channel_read(struct dt_device_node *scmi_node, > + struct scmi_data *scmi_data, > + u64 *addr) > +{ > + u64 size; > + > + if ( !dt_property_read_u32(scmi_node, "arm,smc-id", &scmi_data->func_id) ) > + { > + printk(XENLOG_ERR "scmi: unable to read smc-id from DT\n"); > + return -ENOENT; > + } > + > + return scmi_dt_read_channel_addr(scmi_node, addr, &size); > +} > + > +static __init int scmi_probe(struct dt_device_node *scmi_node, const void *data) > +{ > + u64 addr; > + int ret; > + struct scmi_channel *channel; > + unsigned int n_agents; > + scmi_msg_header_t hdr; > + struct scmi_msg_base_attributes_p2a rx; > + > + ASSERT(scmi_node != NULL); > + > + /* > + * Only bind to the SCMI node provided by Xen under the xen,sci container > + * (e.g. /chosen/xen/xen_scmi_config/scmi). This avoids binding to firmware > + * SCMI nodes that belong to the host OSPM and keeps the mediator scoped to > + * Xen-provided configuration only. > + */ > + if ( !scmi_is_under_xen_sci(scmi_node) ) > + return -ENODEV; > + > + if ( !acpi_disabled ) > + { > + printk(XENLOG_WARNING "scmi: is not supported when using ACPI\n"); > + return -EINVAL; > + } > + > + ret = scmi_check_smccc_ver(); > + if ( ret ) > + return ret; > + > + ret = scmi_dt_hyp_channel_read(scmi_node, &scmi_data, &addr); > + if ( ret ) > + return ret; > + > + INIT_LIST_HEAD(&scmi_data.channel_list); > + spin_lock_init(&scmi_data.channel_list_lock); > + > + scmi_data.dt_dev = scmi_node; > + > + channel = smc_create_channel(SCMI_BASE_AGENT_ID_OWN, scmi_data.func_id, addr); > + if ( IS_ERR(channel) ) > + { > + ret = PTR_ERR(channel); > + goto out; > + } > + > + /* Mark as Xen management channel before collecting agent ID */ > + channel->domain_id = DOMID_XEN; > + > + /* Request agent id for Xen management channel */ > + ret = collect_agent_id(channel); > + if ( ret ) > + goto error; > + > + /* Save the agent id for Xen management channel */ > + scmi_data.hyp_channel_agent_id = channel->agent_id; > + > + hdr.id = SCMI_BASE_PROTOCOL_ATTIBUTES; > + hdr.type = 0; > + hdr.protocol = SCMI_BASE_PROTOCOL; > + > + ret = do_smc_xfer(channel, &hdr, NULL, 0, &rx, sizeof(rx)); > + if ( ret ) > + goto error; > + > + n_agents = SCMI_FIELD_GET(SCMI_BASE_ATTR_NUM_AGENT, rx.attributes); > + printk(XENLOG_DEBUG "scmi: Got agent count %d\n", n_agents); > + ret = collect_agents(scmi_node); > + if ( ret ) > + goto error; > + > + ret = sci_register(&scmi_ops); > + if ( ret ) > + { > + printk(XENLOG_ERR "SCMI: mediator already registered (ret = %d)\n", > + ret); > + goto error; > + } > + > + scmi_data.initialized = true; > + goto out; > + > +error: > + unmap_channel_memory(channel); > + free_channel_list(); > +out: > + return ret; > +} > + > +static const struct dt_device_match scmi_smc_match[] __initconst = { > + DT_MATCH_COMPATIBLE("arm,scmi-smc"), > + { /* sentinel */ }, > +}; > + > +DT_DEVICE_START(scmi_smc_ma, "SCMI SMC MEDIATOR", DEVICE_FIRMWARE) > + .dt_match = scmi_smc_match, > + .init = scmi_probe, > +DT_DEVICE_END > + > +/* > + * Local variables: > + * mode: C > + * c-file-style: "BSD" > + * c-basic-offset: 4 > + * tab-width: 4 > + * indent-tabs-mode: nil > + * End: > + */ > diff --git a/xen/include/public/arch-arm.h b/xen/include/public/arch-arm.h > index 14d51e0746..722c2d2e3a 100644 > --- a/xen/include/public/arch-arm.h > +++ b/xen/include/public/arch-arm.h > @@ -329,6 +329,7 @@ DEFINE_XEN_GUEST_HANDLE(vcpu_guest_context_t); > > #define XEN_DOMCTL_CONFIG_ARM_SCI_NONE 0 > #define XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC 1 > +#define XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC_MA 2 > > #define XEN_DOMCTL_CONFIG_ARM_V8R_EL1_MSA_NONE 0 > #define XEN_DOMCTL_CONFIG_ARM_V8R_EL1_MSA_PMSA 1 > @@ -347,7 +348,9 @@ struct xen_arch_domainconfig { > uint8_t arm_sci_type; > /* IN */ > uint8_t v8r_el1_msa; > - uint16_t pad; > + /* IN */ > + uint8_t arm_sci_agent_id; > + uint8_t pad; > }; > #endif /* __XEN__ || __XEN_TOOLS__ */ -- WBR, Volodymyr ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v14 4/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent driver 2026-10-23 1:56 ` Volodymyr Babchuk @ 2026-10-26 15:39 ` Oleksii Moisieiev 2026-10-29 18:55 ` Volodymyr Babchuk 0 siblings, 1 reply; 10+ messages in thread From: Oleksii Moisieiev @ 2026-10-26 15:39 UTC (permalink / raw) To: Volodymyr Babchuk Cc: xen-devel@lists.xenproject.org, Andrew Cooper, Anthony PERARD, Bertrand Marquis, Jan Beulich, Juergen Gross, Julien Grall, Michal Orzel, Roger Pau Monné, Stefano Stabellini, Grygorii Strashko Hi Volodymyr, Thank you for the review. I agree with the most of your comments, just few more questions left: On 23/10/2026 04:56, Volodymyr Babchuk wrote: > Hi Oleksii, > > Oleksii Moisieiev <Oleksii_Moisieiev@epam.com> writes: > >> This patch introduces SCI driver to support for ARM EL3 Trusted Firmware-A >> (TF-A) which provides SCMI interface with multi-agent support, as shown [snip] >> diff --git a/xen/arch/arm/firmware/scmi-shmem.c b/xen/arch/arm/firmware/scmi-shmem.c >> new file mode 100644 >> index 0000000000..e36745a85e >> --- /dev/null >> +++ b/xen/arch/arm/firmware/scmi-shmem.c >> @@ -0,0 +1,118 @@ >> +/* SPDX-License-Identifier: GPL-2.0-only */ >> +/* >> + * SMC/HVC shmem transport implementation used by >> + * SCI SCMI multi-agent driver. >> + * >> + * Oleksii Moisieiev <oleksii_moisieiev@epam.com> >> + * Copyright (c) 2025 EPAM Systems >> + */ >> + >> +#include <xen/err.h> >> +#include <xen/io.h> >> +#include <asm/io.h> >> + >> +#include "scmi-proto.h" >> +#include "scmi-shmem.h" >> + >> +static inline int >> +shmem_channel_status(const volatile struct scmi_shared_mem __iomem *shmem) >> +{ >> + return (readl(&shmem->channel_status) & >> + SCMI_SHMEM_CHAN_STAT_CHANNEL_FREE) ? 0 : -EBUSY; >> +} >> + >> +int shmem_put_message(volatile struct scmi_shared_mem __iomem *shmem, >> + scmi_msg_header_t *hdr, void *data, unsigned int len) >> +{ >> + int ret; >> + >> + if ( (len + offsetof(struct scmi_shared_mem, msg_payload)) > >> + SCMI_SHMEM_MAPPED_SIZE ) >> + { >> + printk(XENLOG_ERR "scmi: Wrong size of smc message. Data is invalid\n"); > "of the SMC message". Also, looks like it is over 80 characters. agreed with both. thank you. >> + return -EINVAL; >> + } >> + >> + ret = shmem_channel_status(shmem); >> + if ( ret ) >> + return ret; >> + >> + writel_relaxed(0x0, &shmem->channel_status); >> + /* Writing 0x0 right now, but "shmem"_FLAG_INTR_ENABLED can be set */ >> + writel_relaxed(0x0, &shmem->flags); >> + writel_relaxed(sizeof(shmem->msg_header) + len, &shmem->length); >> + writel(pack_scmi_header(hdr), &shmem->msg_header); > Why do you mix writel() and writel_relaxed() here? The main idea that I'm filling up SCMI header information following Device memory ordering and execute __iowmb before store in the last writel call. >> + >> + if ( len > 0 && data ) >> + memcpy_toio(shmem->msg_payload, data, len); >> + >> + return 0; >> +} >> + >> +int shmem_get_response(const volatile struct scmi_shared_mem __iomem *shmem, >> + scmi_msg_header_t *hdr, void *data, unsigned int len) >> +{ >> + int recv_len; >> + int ret; >> + /* >> + * First word of msg_payload carries the returned status; exclude it from >> + * recv_len so only the protocol payload is copied back to the caller. >> + */ >> + int pad = sizeof(hdr->status); >> + >> + if ( len >= SCMI_SHMEM_MAPPED_SIZE - >> + offsetof(struct scmi_shared_mem, msg_payload) ) >> + { >> + printk(XENLOG_ERR >> + "scmi: Wrong size of input smc message. Data may be invalid\n"); > "of the input SMC message" +1 >> + return -EINVAL; >> + } >> + >> + ret = shmem_channel_status(shmem); >> + if ( ret ) >> + return ret; >> + >> + recv_len = readl(&shmem->length) - sizeof(shmem->msg_header); >> + >> + if ( recv_len < 0 ) >> + { >> + printk(XENLOG_ERR >> + "scmi: Wrong size of smc message. Data may be invalid\n"); > "of the SMC message" > > Also, I am wondering if you really need to print this error messages in > the first place. > These generic messages were very helpful in determining the exact error. For example, this function returns -EINVAL both if len >= SCMI_SHMEM_MAPPED_SIZE - offsetof(struct scmi_shared_mem, msg_payload) and if recv_len < 0 However, if you think they are excessive, I can remove the prints and return different error codes instead. >> + return -EINVAL; >> + } >> + >> + unpack_scmi_header(readl(&shmem->msg_header), hdr); >> + >> + hdr->status = readl(&shmem->msg_payload); >> + recv_len = recv_len > pad ? recv_len - pad : 0; >> + >> + ret = scmi_to_xen_errno(hdr->status); >> + if ( ret ) >> + { >> + printk(XENLOG_DEBUG "scmi: Error received: %d\n", ret); >> + return ret; >> + } >> + >> + if ( recv_len > len ) >> + { >> + printk(XENLOG_ERR >> + "scmi: Not enough buffer for message %d, expecting %d\n", >> + recv_len, len); >> + return -EINVAL; >> + } >> + >> + if ( recv_len > 0 ) >> + memcpy_fromio(data, shmem->msg_payload + pad, recv_len); >> + >> + return 0; >> +} >> + >> +/* >> + * Local variables: >> + * mode: C >> + * c-file-style: "BSD" >> + * c-basic-offset: 4 >> + * tab-width: 4 >> + * indent-tabs-mode: nil >> + * End: >> + */ >> diff --git a/xen/arch/arm/firmware/scmi-shmem.h b/xen/arch/arm/firmware/scmi-shmem.h >> new file mode 100644 >> index 0000000000..722263aa77 >> --- /dev/null >> +++ b/xen/arch/arm/firmware/scmi-shmem.h >> @@ -0,0 +1,45 @@ >> +/* SPDX-License-Identifier: GPL-2.0-only */ >> +/* >> + * Arm System Control and Management Interface definitions >> + * Version 3.0 (DEN0056C) >> + * Shared Memory based Transport >> + * >> + * Copyright (c) 2024 EPAM Systems >> + */ >> + >> +#ifndef ARM_FIRMWARE_SCMI_SHMEM_H_ >> +#define ARM_FIRMWARE_SCMI_SHMEM_H_ >> + >> +#include <xen/stdint.h> >> + >> +#define SCMI_SHMEM_CHAN_STAT_CHANNEL_FREE BIT(0, UL) >> +#define SCMI_SHMEM_CHAN_STAT_CHANNEL_ERROR BIT(1, UL) >> + >> +struct scmi_shared_mem { >> + uint32_t reserved; >> + uint32_t channel_status; >> + uint32_t reserved1[2]; >> + uint32_t flags; >> + uint32_t length; >> + uint32_t msg_header; >> + uint8_t msg_payload[]; >> +}; > Should it be __packed? The layout already matches the shared-memory format without padding: Every fixed member is a uint32_t or an array of uint8_t, and the mapping is page aligned. Therefore __packed would not change these offsets, but would reduce the structure's alignment guarantee. I would prefer to keep the natural alignment and add compile-time offset checks to document the transport layout. Does that address your concern? >> + >> +#define SCMI_SHMEM_MAPPED_SIZE PAGE_SIZE >> + >> +int shmem_put_message(volatile struct scmi_shared_mem __iomem *shmem, > I am not sure if volatile makes sense here, taking into account that > this is __iomem anyway. > > Also, shmem_put_message sounds to generic... Maybe scmi_put_message? Agreed. volatile looks unneeded here. I will remove volatile and use scmi_{get/put}_message function names. >> + scmi_msg_header_t *hdr, void *data, unsigned int len); >> + >> +int shmem_get_response(const volatile struct scmi_shared_mem __iomem *shmem, >> + scmi_msg_header_t *hdr, void *data, unsigned int len); > Basically the same comments as for previous one. + >> +#endif /* ARM_FIRMWARE_SCMI_SHMEM_H_ */ >> + >> +/* >> + * Local variables: >> + * mode: C >> + * c-file-style: "BSD" >> + * c-basic-offset: 4 >> + * tab-width: 4 >> + * indent-tabs-mode: nil >> + * End: >> + */ >> diff --git a/xen/arch/arm/firmware/scmi-smc-multiagent.c b/xen/arch/arm/firmware/scmi-smc-multiagent.c >> new file mode 100644 >> index 0000000000..6c46f6cd36 >> --- /dev/null >> +++ b/xen/arch/arm/firmware/scmi-smc-multiagent.c >> @@ -0,0 +1,792 @@ >> +/* SPDX-License-Identifier: GPL-2.0-only */ >> +/* >> + * SCI SCMI multi-agent driver, using SMC/HVC shmem as transport. >> + * >> + * Oleksii Moisieiev <oleksii_moisieiev@epam.com> >> + * Copyright (c) 2025 EPAM Systems >> + */ >> + >> +#include <xen/acpi.h> >> + >> +#include <xen/device_tree.h> >> +#include <xen/init.h> >> +#include <xen/iocap.h> >> +#include <xen/err.h> >> +#include <xen/libfdt/libfdt.h> >> +#include <xen/string.h> >> +#include <xen/param.h> >> +#include <xen/sched.h> >> +#include <xen/vmap.h> >> + >> +#include <asm/firmware/sci.h> >> +#include <asm/smccc.h> >> + >> +#include "scmi-proto.h" >> +#include "scmi-shmem.h" >> + >> +struct scmi_channel { >> + uint32_t agent_id; > If all agent_ids large than UINT8_MAX are invalid, why do you need > uint32_t here? According to the DEN0056F document: https://support.arm.com/documentation/den0056/f/?lang=en agent_id is uint32_t in SCMI messages. >> + uint32_t func_id; >> + domid_t domain_id; >> + uint64_t paddr; > paddr_t +1. thanks. >> + struct scmi_shared_mem __iomem *shmem; >> + spinlock_t lock; >> + struct list_head list; >> +}; >> + >> +struct scmi_data { >> + struct list_head channel_list; >> + spinlock_t channel_list_lock; >> + uint32_t func_id; >> + bool initialized; >> + uint32_t shmem_phandle; >> + uint32_t hyp_channel_agent_id; >> + struct dt_device_node *dt_dev; >> +}; >> + >> +static struct scmi_data scmi_data; >> + >> +static bool scmi_is_under_xen_sci(const struct dt_device_node *node) >> +{ >> + const struct dt_device_node *p; >> + >> + for ( p = node->parent; p; p = p->parent ) >> + if ( dt_device_is_compatible(p, "xen,sci") ) >> + return true; >> + >> + return false; >> +} >> + >> +static int send_smc_message(struct scmi_channel *chan_info, >> + scmi_msg_header_t *hdr, void *data, int len) >> +{ >> + struct arm_smccc_res resp; >> + int ret; >> + >> + ret = shmem_put_message(chan_info->shmem, hdr, data, len); >> + if ( ret ) >> + return ret; >> + >> + resp = arm_smccc_1_1_smc(chan_info->func_id, 0, 0, 0, 0, 0, 0, 0); >> + >> + if ( resp.a0 == (unsigned long)ARM_SMCCC_INVALID_PARAMETER ) >> + return -EINVAL; >> + >> + if ( resp.a0 ) >> + return -EOPNOTSUPP; >> + >> + return 0; >> +} >> + >> +static int do_smc_xfer(struct scmi_channel *chan_info, scmi_msg_header_t *hdr, >> + void *tx_data, int tx_size, void *rx_data, int rx_size) >> +{ >> + int ret = 0; >> + >> + ASSERT(chan_info && chan_info->shmem); >> + >> + if ( !hdr ) >> + return -EINVAL; >> + >> + spin_lock(&chan_info->lock); >> + >> + printk(XENLOG_DEBUG >> + "scmi: agent_id = %d msg_id = %x type = %d, proto = %x\n", >> + chan_info->agent_id, hdr->id, hdr->type, hdr->protocol); >> + >> + ret = send_smc_message(chan_info, hdr, tx_data, tx_size); >> + if ( ret ) >> + goto clean; >> + >> + ret = shmem_get_response(chan_info->shmem, hdr, rx_data, rx_size); >> + >> +clean: >> + printk(XENLOG_DEBUG >> + "scmi: get smc response agent_id = %d msg_id = %x proto = %x res=%d\n", >> + chan_info->agent_id, hdr->id, hdr->protocol, ret); > You will print this even if send_smc_message() fails. I think it should > go before "clean:" label. > > Also, maybe "cleanup:" ? +1. It's better to place debug print before "clean" lavel. >> + >> + spin_unlock(&chan_info->lock); >> + >> + return ret; >> +} >> + >> +static struct scmi_channel *get_channel_by_id(uint32_t agent_id) >> +{ >> + struct scmi_channel *curr; >> + bool found = false; >> + >> + spin_lock(&scmi_data.channel_list_lock); >> + list_for_each_entry(curr, &scmi_data.channel_list, list) >> + { >> + if ( curr->agent_id == agent_id ) >> + { >> + found = true; >> + break; >> + } >> + } >> + >> + spin_unlock(&scmi_data.channel_list_lock); >> + if ( found ) >> + return curr; >> + >> + return NULL; >> +} >> + >> +static struct scmi_channel *acquire_scmi_channel(struct domain *d, >> + uint32_t agent_id) >> +{ >> + struct scmi_channel *curr; >> + struct scmi_channel *ret = ERR_PTR(-ENOENT); >> + >> + spin_lock(&scmi_data.channel_list_lock); >> + list_for_each_entry(curr, &scmi_data.channel_list, list) >> + { >> + if ( curr->agent_id == agent_id ) >> + { >> + if ( curr->domain_id != DOMID_INVALID ) >> + { >> + ret = ERR_PTR(-EEXIST); >> + break; >> + } >> + >> + curr->domain_id = d->domain_id; >> + ret = curr; >> + break; >> + } >> + } >> + >> + spin_unlock(&scmi_data.channel_list_lock); >> + >> + return ret; >> +} >> + >> +static void relinquish_scmi_channel(struct scmi_channel *channel) >> +{ >> + ASSERT(channel != NULL); >> + >> + spin_lock(&scmi_data.channel_list_lock); >> + channel->domain_id = DOMID_INVALID; >> + spin_unlock(&scmi_data.channel_list_lock); >> +} >> + >> +static int map_channel_memory(struct scmi_channel *channel) >> +{ >> + ASSERT(channel && channel->paddr); >> + channel->shmem = ioremap_nocache(channel->paddr, SCMI_SHMEM_MAPPED_SIZE); >> + if ( !channel->shmem ) >> + return -ENOMEM; >> + >> + channel->shmem->channel_status = SCMI_SHMEM_CHAN_STAT_CHANNEL_FREE; >> + printk(XENLOG_DEBUG "scmi: Got shmem %lx after vmap %p\n", channel->paddr, > You will need proper format, not "%lx" for channel->paddr Agree. with paddr_t I will use PRIpaddr format. >> + channel->shmem); >> + >> + return 0; >> +} >> + >> +static void unmap_channel_memory(struct scmi_channel *channel) >> +{ >> + ASSERT(channel); >> + >> + if ( !channel->shmem ) >> + return; >> + >> + iounmap(channel->shmem); >> + channel->shmem = NULL; >> +} >> + >> +static struct scmi_channel *smc_create_channel(uint32_t agent_id, >> + uint32_t func_id, uint64_t addr) >> +{ >> + struct scmi_channel *channel, *curr; >> + >> + spin_lock(&scmi_data.channel_list_lock); >> + >> + /* Check if channel already exists while holding the lock */ >> + list_for_each_entry(curr, &scmi_data.channel_list, list) >> + { >> + if ( curr->agent_id == agent_id ) >> + { >> + spin_unlock(&scmi_data.channel_list_lock); >> + return ERR_PTR(-EEXIST); >> + } >> + } >> + >> + channel = xmalloc(struct scmi_channel); >> + if ( !channel ) >> + { >> + spin_unlock(&scmi_data.channel_list_lock); >> + return ERR_PTR(-ENOMEM); >> + } >> + >> + spin_lock_init(&channel->lock); >> + channel->agent_id = agent_id; >> + channel->func_id = func_id; >> + channel->domain_id = DOMID_INVALID; >> + channel->shmem = NULL; >> + channel->paddr = addr; >> + list_add_tail(&channel->list, &scmi_data.channel_list); >> + >> + spin_unlock(&scmi_data.channel_list_lock); >> + return channel; >> +} >> + >> +static void free_channel_list(void) >> +{ >> + struct scmi_channel *curr, *_curr; >> + /* >> + * Called only on the __init error path, before any runtime users exist, > Then you can put this function into __init as well +1 >> + * so no other thread is iterating channel_list. Keep the lock for the >> + * whole drain for clarity and to avoid misleading readers about possible >> + * concurrent access. >> + */ >> + spin_lock(&scmi_data.channel_list_lock); >> + list_for_each_entry_safe(curr, _curr, &scmi_data.channel_list, list) >> + { >> + list_del(&curr->list); >> + xfree(curr); >> + } >> + spin_unlock(&scmi_data.channel_list_lock); >> +} >> + >> +static int __init >> +scmi_dt_read_channel_addr(struct dt_device_node *scmi_node, u64 *addr, >> + u64 *size) >> +{ >> + struct dt_device_node *shmem_node; >> + uint32_t phandle; >> + int ret; >> + >> + ret = dt_property_read_variable_u32_array(scmi_node, "shmem", &phandle, >> + 1, 1); >> + if ( ret < 0 ) >> + return ret; >> + >> + shmem_node = dt_find_node_by_phandle(phandle); >> + if ( !shmem_node || >> + !dt_device_is_compatible(shmem_node, "arm,scmi-shmem") ) >> + return -EINVAL; >> + >> + ret = dt_device_get_address(shmem_node, 0, addr, size); >> + if ( ret ) >> + return ret; >> + >> + if ( !*size || !IS_ALIGNED(*size, SCMI_SHMEM_MAPPED_SIZE) || >> + !IS_ALIGNED(*addr, SCMI_SHMEM_MAPPED_SIZE) ) >> + { >> + printk(XENLOG_ERR "scmi: invalid shmem range for %s\n", >> + scmi_node->full_name); >> + return -EINVAL; >> + } >> + >> + return 0; >> +} >> + >> +/* >> + * Handle Dom0 SCMI specific DT nodes >> + * >> + * Make a decision on copying SCMI specific nodes into Dom0 device tree. >> + * For SCMI multi-agent case: >> + * - shmem nodes will not be copied and generated instead if SCMI >> + * is enabled for Dom0 >> + * - scmi node will be copied if SCMI is enabled for Dom0 >> + */ >> +static bool scmi_dt_handle_node(struct domain *d, struct dt_device_node *node) > Also __init? + >> +{ >> + static const struct dt_device_match shmem_matches[] __initconst = { >> + DT_MATCH_COMPATIBLE("arm,scmi-shmem"), >> + { /* sentinel */ }, >> + }; >> + static const struct dt_device_match scmi_matches[] __initconst = { >> + DT_MATCH_PATH("/firmware/scmi"), >> + { /* sentinel */ }, >> + }; >> + >> + if ( !scmi_data.initialized ) >> + return false; >> + >> + /* skip scmi shmem node for dom0 if scmi not enabled */ >> + if ( dt_match_node(shmem_matches, node) && !sci_domain_is_enabled(d) ) >> + { >> + dt_dprintk(" Skip scmi shmem node\n"); >> + return true; >> + } >> + >> + /* drop scmi if not enabled */ >> + if ( dt_match_node(scmi_matches, node) && !sci_domain_is_enabled(d) ) >> + { >> + dt_dprintk(" Skip scmi node\n"); >> + return true; >> + } >> + >> + return false; >> +} >> + >> +static int scmi_assign_device(uint32_t agent_id, uint32_t device_id, >> + uint32_t flags) >> +{ >> + struct scmi_msg_base_set_device_permissions_a2p tx; >> + struct scmi_channel *channel; >> + scmi_msg_header_t hdr; >> + >> + channel = get_channel_by_id(scmi_data.hyp_channel_agent_id); >> + if ( !channel ) >> + return -EINVAL; >> + >> + hdr.id = SCMI_BASE_SET_DEVICE_PERMISSIONS; >> + hdr.type = 0; >> + hdr.protocol = SCMI_BASE_PROTOCOL; >> + >> + tx.agent_id = agent_id; >> + tx.device_id = device_id; >> + tx.flags = flags; >> + >> + return do_smc_xfer(channel, &hdr, &tx, sizeof(tx), NULL, 0); >> +} >> + >> +static int scmi_dt_assign_device(struct domain *d, >> + struct dt_phandle_args *ac_spec) >> +{ >> + struct scmi_channel *agent_channel; >> + uint32_t scmi_device_id = ac_spec->args[0]; >> + int ret; >> + >> + if ( !d->arch.sci_data ) >> + return 0; >> + >> + /* The access-controllers is specified for DT dev, but it's not a SCMI */ >> + if ( !scmi_data.dt_dev || >> + !dt_node_path_is_equal(ac_spec->np, scmi_data.dt_dev->full_name) ) >> + return 0; >> + >> + agent_channel = d->arch.sci_data; >> + >> + spin_lock(&agent_channel->lock); >> + >> + ret = scmi_assign_device(agent_channel->agent_id, scmi_device_id, >> + SCMI_BASE_DEVICE_ACCESS_ALLOW); >> + if ( ret ) >> + { >> + printk(XENLOG_ERR >> + "scmi: could not assign dev for %pd agent:%d dev_id:%u (%d)", >> + d, agent_channel->agent_id, scmi_device_id, ret); >> + } >> + >> + spin_unlock(&agent_channel->lock); >> + return ret; >> +} >> + >> +static int collect_agent_id(struct scmi_channel *agent_channel) >> +{ >> + int ret; >> + scmi_msg_header_t hdr; >> + struct scmi_msg_base_discover_agent_p2a da_rx; >> + struct scmi_msg_base_discover_agent_a2p da_tx; >> + >> + ret = map_channel_memory(agent_channel); >> + if ( ret ) >> + return ret; >> + >> + hdr.id = SCMI_BASE_DISCOVER_AGENT; >> + hdr.type = 0; >> + hdr.protocol = SCMI_BASE_PROTOCOL; >> + >> + da_tx.agent_id = agent_channel->agent_id; >> + >> + ret = do_smc_xfer(agent_channel, &hdr, &da_tx, sizeof(da_tx), &da_rx, >> + sizeof(da_rx)); >> + if ( agent_channel->domain_id != DOMID_XEN ) >> + unmap_channel_memory(agent_channel); >> + if ( ret ) > On error, Xen's channel will be mapped still. Is this intended behaviour? It remains mapped when collect_agent_id() returns, but the existing caller cleans it up. scmi_probe() marks the management channel DOMID_XEN before discovery. If discovery fails, it jumps to error, calls unmap_channel_memory(channel), then free_channel_list(). Thus the current probe failure path does not leak the mapping. On successful discovery the management mapping is deliberately retained for subsequent permission and reset requests. Secondary discovery mappings are always removed by collect_agent_id(), whether the transfer succeeds or fails. >> + return ret; >> + >> + printk(XENLOG_DEBUG "id=0x%x name=%s\n", da_rx.agent_id, da_rx.name); >> + agent_channel->agent_id = da_rx.agent_id; >> + return 0; >> +} >> + >> +static int __init collect_agents(struct dt_device_node *scmi_node) >> +{ >> + struct dt_device_node *agents, *node; >> + unsigned int count = 0; >> + >> + /* Use the configuration containing this management channel. */ >> + for ( agents = scmi_node->parent; agents; agents = agents->parent ) >> + if ( dt_device_is_compatible(agents, "xen,sci") ) >> + break; >> + >> + ASSERT(agents); >> + >> + dt_for_each_child_node(agents, node) >> + if ( dt_node_name_is_equal(node, "agents") ) >> + break; >> + >> + if ( !node ) >> + { >> + printk(XENLOG_ERR "scmi: missing agents node\n"); >> + return -EINVAL; >> + } >> + agents = node; >> + >> + if ( dt_child_n_addr_cells(agents) != 1 || >> + dt_child_n_size_cells(agents) != 0 ) >> + return -EINVAL; >> + >> + dt_for_each_child_node(agents, node) >> + { >> + struct scmi_channel agent = { >> + .agent_id = SCMI_BASE_AGENT_ID_OWN, >> + .domain_id = DOMID_INVALID, >> + }; >> + struct scmi_channel *channel; >> + bool discover; >> + uint32_t len; >> + u64 size; > uint64_t > > +1 >> + int ret; >> + >> + if ( !dt_device_is_compatible(node, "xen,scmi-agent") || >> + !dt_device_is_available(node) ) >> + continue; >> + >> + discover = !dt_find_property(node, "reg", &len); >> + if ( !discover ) >> + { >> + if ( len != sizeof(uint32_t) || >> + !dt_property_read_u32(node, "reg", &agent.agent_id) ) >> + return -EINVAL; >> + } >> + >> + if ( !dt_property_read_u32(node, "arm,smc-id", &agent.func_id) ) >> + return -EINVAL; >> + >> + ret = scmi_dt_read_channel_addr(node, &agent.paddr, &size); >> + if ( ret ) >> + return ret; >> + >> + if ( discover ) >> + { >> + /* Resolve the ID before inserting it into the channel list. */ >> + spin_lock_init(&agent.lock); >> + ret = collect_agent_id(&agent); >> + if ( ret ) >> + return ret; >> + } >> + >> + /* Guest configurations encode agent IDs in an eight-bit field. */ >> + if ( agent.agent_id >= UINT8_MAX ) >> + return -EINVAL; >> + >> + channel = smc_create_channel(agent.agent_id, agent.func_id, >> + agent.paddr); >> + if ( IS_ERR(channel) ) >> + { >> + printk(XENLOG_ERR "scmi: Could not create channel for %s: %ld\n", >> + node->full_name, PTR_ERR(channel)); >> + return PTR_ERR(channel); >> + } >> + >> + count++; >> + printk(XENLOG_DEBUG "scmi: Agent %u SMC %X addr %lx\n", >> + agent.agent_id, agent.func_id, (unsigned long)agent.paddr); > Why cast to unsigned long? Was used to match %lx. Using PRIpaddr will drop the need of casing. >> + } >> + >> + return count ? 0 : -EINVAL; >> +} >> + >> +static int scmi_domain_init(struct domain *d, >> + struct xen_domctl_createdomain *config) >> +{ >> + struct scmi_channel *channel; >> + int ret; >> + >> + if ( !scmi_data.initialized ) >> + return 0; >> + >> + /* >> + * SCMI support is configured via: >> + * - For dom0: xen,dom0-sci-agent-id property under the xen,sci container >> + * - For dom0less: xen,sci-agent-id in the domain node >> + * The config->arch.arm_sci_type and config->arch.arm_sci_agent_id >> + * are already set by domain_build.c or dom0less-build.c >> + */ >> + >> + if ( config->arch.arm_sci_type == XEN_DOMCTL_CONFIG_ARM_SCI_NONE ) >> + return 0; >> + >> + channel = acquire_scmi_channel(d, config->arch.arm_sci_agent_id); >> + if ( IS_ERR(channel) ) >> + { >> + printk(XENLOG_ERR >> + "scmi: Failed to acquire SCMI channel for agent_id %u: %ld\n", >> + config->arch.arm_sci_agent_id, PTR_ERR(channel)); >> + return PTR_ERR(channel); >> + } >> + >> + printk(XENLOG_INFO >> + "scmi: Acquire channel id = 0x%x, domain_id = %d paddr = 0x%lx\n", >> + channel->agent_id, channel->domain_id, channel->paddr); >> + >> + /* >> + * Dom0 (if present) needs to have an access to the guest memory range >> + * to satisfy iomem_access_permitted() check in XEN_DOMCTL_iomem_permission >> + * domctl. >> + */ >> + if ( hardware_domain && !is_hardware_domain(d) ) >> + { >> + ret = iomem_permit_access(hardware_domain, paddr_to_pfn(channel->paddr), >> + paddr_to_pfn(channel->paddr + >> + SCMI_SHMEM_MAPPED_SIZE - 1)); >> + if ( ret ) >> + goto error; >> + } >> + >> + d->arch.sci_data = channel; >> + d->arch.sci_enabled = true; >> + >> + return 0; >> + >> +error: >> + relinquish_scmi_channel(channel); >> + return ret; >> +} >> + >> +int scmi_domain_sanitise_config(const struct xen_domctl_createdomain *config) >> +{ >> + if ( config->arch.arm_sci_type != XEN_DOMCTL_CONFIG_ARM_SCI_NONE && >> + config->arch.arm_sci_type != XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC_MA ) >> + { >> + dprintk(XENLOG_INFO, "scmi: Unsupported ARM_SCI type\n"); >> + return -EINVAL; >> + } >> + >> + return 0; >> +} >> + >> +static int scmi_relinquish_resources(struct domain *d) >> +{ >> + int ret; >> + struct scmi_channel *channel, *agent_channel; >> + scmi_msg_header_t hdr; >> + struct scmi_msg_base_reset_agent_cfg_a2p tx; >> + >> + if ( !d->arch.sci_data ) >> + return 0; >> + >> + agent_channel = d->arch.sci_data; >> + >> + spin_lock(&agent_channel->lock); >> + tx.agent_id = agent_channel->agent_id; >> + spin_unlock(&agent_channel->lock); >> + >> + channel = get_channel_by_id(scmi_data.hyp_channel_agent_id); >> + if ( !channel ) >> + { >> + printk(XENLOG_ERR >> + "scmi: Unable to get Hypervisor scmi channel for domain %d\n", >> + d->domain_id); >> + return -EINVAL; >> + } >> + >> + hdr.id = SCMI_BASE_RESET_AGENT_CONFIGURATION; >> + hdr.type = 0; >> + hdr.protocol = SCMI_BASE_PROTOCOL; >> + >> + tx.flags = SCMI_BASE_AGENT_PERMISSIONS_RESET; >> + >> + ret = do_smc_xfer(channel, &hdr, &tx, sizeof(tx), NULL, 0); >> + if ( ret == -EOPNOTSUPP ) >> + return 0; >> + >> + return ret; >> +} >> + >> +static void scmi_domain_destroy(struct domain *d) >> +{ >> + struct scmi_channel *channel; >> + >> + if ( !d->arch.sci_data ) >> + return; >> + >> + channel = d->arch.sci_data; >> + spin_lock(&channel->lock); >> + >> + relinquish_scmi_channel(channel); >> + printk(XENLOG_DEBUG "scmi: Free domain %d\n", d->domain_id); > You can use "%pd" + >> + >> + d->arch.sci_data = NULL; >> + d->arch.sci_enabled = false; >> + >> + spin_unlock(&channel->lock); >> +} >> + >> +static bool scmi_handle_call(struct cpu_user_regs *regs) >> +{ >> + uint32_t fid = (uint32_t)get_user_reg(regs, 0); >> + struct scmi_channel *agent_channel; >> + struct domain *d = current->domain; >> + bool res = false; >> + >> + if ( (!sci_domain_is_enabled(d)) || (!d->arch.sci_data) ) >> + return false; >> + >> + agent_channel = d->arch.sci_data; >> + spin_lock(&agent_channel->lock); >> + >> + if ( agent_channel->func_id != fid ) >> + { >> + res = false; >> + goto unlock; >> + } >> + >> + arm_smccc_guest_smc(regs); >> + res = true; >> +unlock: >> + spin_unlock(&agent_channel->lock); >> + >> + return res; >> +} >> + >> +static const struct sci_mediator_ops scmi_ops = { >> + .domain_init = scmi_domain_init, >> + .domain_destroy = scmi_domain_destroy, >> + .relinquish_resources = scmi_relinquish_resources, >> + .handle_call = scmi_handle_call, >> + .dom0_dt_handle_node = scmi_dt_handle_node, >> + .domain_sanitise_config = scmi_domain_sanitise_config, >> + .assign_dt_device = scmi_dt_assign_device, >> +}; >> + >> +static int __init scmi_check_smccc_ver(void) >> +{ >> + if ( smccc_ver < ARM_SMCCC_VERSION_1_1 ) >> + { >> + printk(XENLOG_WARNING >> + "scmi: No SMCCC 1.1 support, SCMI calls forwarding disabled\n"); >> + return -ENOSYS; >> + } >> + >> + return 0; >> +} >> + >> +static int __init scmi_dt_hyp_channel_read(struct dt_device_node *scmi_node, >> + struct scmi_data *scmi_data, >> + u64 *addr) >> +{ >> + u64 size; >> + >> + if ( !dt_property_read_u32(scmi_node, "arm,smc-id", &scmi_data->func_id) ) >> + { >> + printk(XENLOG_ERR "scmi: unable to read smc-id from DT\n"); >> + return -ENOENT; >> + } >> + >> + return scmi_dt_read_channel_addr(scmi_node, addr, &size); >> +} >> + >> +static __init int scmi_probe(struct dt_device_node *scmi_node, const void *data) >> +{ >> + u64 addr; >> + int ret; >> + struct scmi_channel *channel; >> + unsigned int n_agents; >> + scmi_msg_header_t hdr; >> + struct scmi_msg_base_attributes_p2a rx; >> + >> + ASSERT(scmi_node != NULL); >> + >> + /* >> + * Only bind to the SCMI node provided by Xen under the xen,sci container >> + * (e.g. /chosen/xen/xen_scmi_config/scmi). This avoids binding to firmware >> + * SCMI nodes that belong to the host OSPM and keeps the mediator scoped to >> + * Xen-provided configuration only. >> + */ >> + if ( !scmi_is_under_xen_sci(scmi_node) ) >> + return -ENODEV; >> + >> + if ( !acpi_disabled ) >> + { >> + printk(XENLOG_WARNING "scmi: is not supported when using ACPI\n"); >> + return -EINVAL; >> + } >> + >> + ret = scmi_check_smccc_ver(); >> + if ( ret ) >> + return ret; >> + >> + ret = scmi_dt_hyp_channel_read(scmi_node, &scmi_data, &addr); >> + if ( ret ) >> + return ret; >> + >> + INIT_LIST_HEAD(&scmi_data.channel_list); >> + spin_lock_init(&scmi_data.channel_list_lock); >> + >> + scmi_data.dt_dev = scmi_node; >> + >> + channel = smc_create_channel(SCMI_BASE_AGENT_ID_OWN, scmi_data.func_id, addr); >> + if ( IS_ERR(channel) ) >> + { >> + ret = PTR_ERR(channel); >> + goto out; >> + } >> + >> + /* Mark as Xen management channel before collecting agent ID */ >> + channel->domain_id = DOMID_XEN; >> + >> + /* Request agent id for Xen management channel */ >> + ret = collect_agent_id(channel); >> + if ( ret ) >> + goto error; >> + >> + /* Save the agent id for Xen management channel */ >> + scmi_data.hyp_channel_agent_id = channel->agent_id; >> + >> + hdr.id = SCMI_BASE_PROTOCOL_ATTIBUTES; >> + hdr.type = 0; >> + hdr.protocol = SCMI_BASE_PROTOCOL; >> + >> + ret = do_smc_xfer(channel, &hdr, NULL, 0, &rx, sizeof(rx)); >> + if ( ret ) >> + goto error; >> + >> + n_agents = SCMI_FIELD_GET(SCMI_BASE_ATTR_NUM_AGENT, rx.attributes); >> + printk(XENLOG_DEBUG "scmi: Got agent count %d\n", n_agents); >> + ret = collect_agents(scmi_node); >> + if ( ret ) >> + goto error; >> + >> + ret = sci_register(&scmi_ops); >> + if ( ret ) >> + { >> + printk(XENLOG_ERR "SCMI: mediator already registered (ret = %d)\n", >> + ret); >> + goto error; >> + } >> + >> + scmi_data.initialized = true; >> + goto out; >> + >> +error: >> + unmap_channel_memory(channel); >> + free_channel_list(); >> +out: >> + return ret; >> +} >> + >> +static const struct dt_device_match scmi_smc_match[] __initconst = { >> + DT_MATCH_COMPATIBLE("arm,scmi-smc"), >> + { /* sentinel */ }, >> +}; >> + >> +DT_DEVICE_START(scmi_smc_ma, "SCMI SMC MEDIATOR", DEVICE_FIRMWARE) >> + .dt_match = scmi_smc_match, >> + .init = scmi_probe, >> +DT_DEVICE_END >> + >> +/* >> + * Local variables: >> + * mode: C >> + * c-file-style: "BSD" >> + * c-basic-offset: 4 >> + * tab-width: 4 >> + * indent-tabs-mode: nil >> + * End: >> + */ >> diff --git a/xen/include/public/arch-arm.h b/xen/include/public/arch-arm.h >> index 14d51e0746..722c2d2e3a 100644 >> --- a/xen/include/public/arch-arm.h >> +++ b/xen/include/public/arch-arm.h >> @@ -329,6 +329,7 @@ DEFINE_XEN_GUEST_HANDLE(vcpu_guest_context_t); >> >> #define XEN_DOMCTL_CONFIG_ARM_SCI_NONE 0 >> #define XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC 1 >> +#define XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC_MA 2 >> >> #define XEN_DOMCTL_CONFIG_ARM_V8R_EL1_MSA_NONE 0 >> #define XEN_DOMCTL_CONFIG_ARM_V8R_EL1_MSA_PMSA 1 >> @@ -347,7 +348,9 @@ struct xen_arch_domainconfig { >> uint8_t arm_sci_type; >> /* IN */ >> uint8_t v8r_el1_msa; >> - uint16_t pad; >> + /* IN */ >> + uint8_t arm_sci_agent_id; >> + uint8_t pad; >> }; >> #endif /* __XEN__ || __XEN_TOOLS__ */ ^ permalink raw reply [flat|nested] 10+ messages in thread
* Re: [PATCH v14 4/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent driver 2026-10-26 15:39 ` Oleksii Moisieiev @ 2026-10-29 18:55 ` Volodymyr Babchuk 0 siblings, 0 replies; 10+ messages in thread From: Volodymyr Babchuk @ 2026-10-29 18:55 UTC (permalink / raw) To: Oleksii Moisieiev Cc: xen-devel@lists.xenproject.org, Andrew Cooper, Anthony PERARD, Bertrand Marquis, Jan Beulich, Juergen Gross, Julien Grall, Michal Orzel, Roger Pau Monné, Stefano Stabellini, Grygorii Strashko Hi Oleksii, Oleksii Moisieiev <Oleksii_Moisieiev@epam.com> writes: > Hi Volodymyr, > > Thank you for the review. I agree with the most of your comments, just > few more questions left: > > On 23/10/2026 04:56, Volodymyr Babchuk wrote: >> Hi Oleksii, >> >> Oleksii Moisieiev <Oleksii_Moisieiev@epam.com> writes: >> >>> This patch introduces SCI driver to support for ARM EL3 Trusted Firmware-A >>> (TF-A) which provides SCMI interface with multi-agent support, as shown > [snip] >>> diff --git a/xen/arch/arm/firmware/scmi-shmem.c b/xen/arch/arm/firmware/scmi-shmem.c >>> new file mode 100644 >>> index 0000000000..e36745a85e >>> --- /dev/null >>> +++ b/xen/arch/arm/firmware/scmi-shmem.c >>> @@ -0,0 +1,118 @@ >>> +/* SPDX-License-Identifier: GPL-2.0-only */ >>> +/* >>> + * SMC/HVC shmem transport implementation used by >>> + * SCI SCMI multi-agent driver. >>> + * >>> + * Oleksii Moisieiev <oleksii_moisieiev@epam.com> >>> + * Copyright (c) 2025 EPAM Systems >>> + */ >>> + >>> +#include <xen/err.h> >>> +#include <xen/io.h> >>> +#include <asm/io.h> >>> + >>> +#include "scmi-proto.h" >>> +#include "scmi-shmem.h" >>> + >>> +static inline int >>> +shmem_channel_status(const volatile struct scmi_shared_mem __iomem *shmem) >>> +{ >>> + return (readl(&shmem->channel_status) & >>> + SCMI_SHMEM_CHAN_STAT_CHANNEL_FREE) ? 0 : -EBUSY; >>> +} >>> + >>> +int shmem_put_message(volatile struct scmi_shared_mem __iomem *shmem, >>> + scmi_msg_header_t *hdr, void *data, unsigned int len) >>> +{ >>> + int ret; >>> + >>> + if ( (len + offsetof(struct scmi_shared_mem, msg_payload)) > >>> + SCMI_SHMEM_MAPPED_SIZE ) >>> + { >>> + printk(XENLOG_ERR "scmi: Wrong size of smc message. Data is invalid\n"); >> "of the SMC message". Also, looks like it is over 80 characters. > agreed with both. thank you. >>> + return -EINVAL; >>> + } >>> + >>> + ret = shmem_channel_status(shmem); >>> + if ( ret ) >>> + return ret; >>> + >>> + writel_relaxed(0x0, &shmem->channel_status); >>> + /* Writing 0x0 right now, but "shmem"_FLAG_INTR_ENABLED can be set */ >>> + writel_relaxed(0x0, &shmem->flags); >>> + writel_relaxed(sizeof(shmem->msg_header) + len, &shmem->length); >>> + writel(pack_scmi_header(hdr), &shmem->msg_header); >> Why do you mix writel() and writel_relaxed() here? > > The main idea that I'm filling up SCMI header information following > Device memory ordering and execute __iowmb before store in the last > writel call. Yes, but difference between writel_relaxed() and writel() is that writel() has memory barrier for Normal Memory accesses. Writes to Device Memory are guaranteed to arrive in the proper order. > > >>> + >>> + if ( len > 0 && data ) >>> + memcpy_toio(shmem->msg_payload, data, len); >>> + >>> + return 0; >>> +} >>> + >>> +int shmem_get_response(const volatile struct scmi_shared_mem __iomem *shmem, >>> + scmi_msg_header_t *hdr, void *data, unsigned int len) >>> +{ >>> + int recv_len; >>> + int ret; >>> + /* >>> + * First word of msg_payload carries the returned status; exclude it from >>> + * recv_len so only the protocol payload is copied back to the caller. >>> + */ >>> + int pad = sizeof(hdr->status); >>> + >>> + if ( len >= SCMI_SHMEM_MAPPED_SIZE - >>> + offsetof(struct scmi_shared_mem, msg_payload) ) >>> + { >>> + printk(XENLOG_ERR >>> + "scmi: Wrong size of input smc message. Data may be invalid\n"); >> "of the input SMC message" > +1 >>> + return -EINVAL; >>> + } >>> + >>> + ret = shmem_channel_status(shmem); >>> + if ( ret ) >>> + return ret; >>> + >>> + recv_len = readl(&shmem->length) - sizeof(shmem->msg_header); >>> + >>> + if ( recv_len < 0 ) >>> + { >>> + printk(XENLOG_ERR >>> + "scmi: Wrong size of smc message. Data may be invalid\n"); >> "of the SMC message" >> >> Also, I am wondering if you really need to print this error messages in >> the first place. >> > These generic messages were very helpful in determining the exact error. > For example, this function returns -EINVAL both if > > len >= SCMI_SHMEM_MAPPED_SIZE - offsetof(struct scmi_shared_mem, msg_payload) and if recv_len < 0 > > However, if you think they are excessive, I can remove the prints and return different error codes instead. Well, I have no strong opinion here. So you can leave them if you wish. Just make sure that they have proper spelling. > >>> + return -EINVAL; >>> + } >>> + >>> + unpack_scmi_header(readl(&shmem->msg_header), hdr); >>> + >>> + hdr->status = readl(&shmem->msg_payload); >>> + recv_len = recv_len > pad ? recv_len - pad : 0; >>> + >>> + ret = scmi_to_xen_errno(hdr->status); >>> + if ( ret ) >>> + { >>> + printk(XENLOG_DEBUG "scmi: Error received: %d\n", ret); >>> + return ret; >>> + } >>> + >>> + if ( recv_len > len ) >>> + { >>> + printk(XENLOG_ERR >>> + "scmi: Not enough buffer for message %d, expecting %d\n", >>> + recv_len, len); >>> + return -EINVAL; >>> + } >>> + >>> + if ( recv_len > 0 ) >>> + memcpy_fromio(data, shmem->msg_payload + pad, recv_len); >>> + >>> + return 0; >>> +} >>> + >>> +/* >>> + * Local variables: >>> + * mode: C >>> + * c-file-style: "BSD" >>> + * c-basic-offset: 4 >>> + * tab-width: 4 >>> + * indent-tabs-mode: nil >>> + * End: >>> + */ >>> diff --git a/xen/arch/arm/firmware/scmi-shmem.h b/xen/arch/arm/firmware/scmi-shmem.h >>> new file mode 100644 >>> index 0000000000..722263aa77 >>> --- /dev/null >>> +++ b/xen/arch/arm/firmware/scmi-shmem.h >>> @@ -0,0 +1,45 @@ >>> +/* SPDX-License-Identifier: GPL-2.0-only */ >>> +/* >>> + * Arm System Control and Management Interface definitions >>> + * Version 3.0 (DEN0056C) >>> + * Shared Memory based Transport >>> + * >>> + * Copyright (c) 2024 EPAM Systems >>> + */ >>> + >>> +#ifndef ARM_FIRMWARE_SCMI_SHMEM_H_ >>> +#define ARM_FIRMWARE_SCMI_SHMEM_H_ >>> + >>> +#include <xen/stdint.h> >>> + >>> +#define SCMI_SHMEM_CHAN_STAT_CHANNEL_FREE BIT(0, UL) >>> +#define SCMI_SHMEM_CHAN_STAT_CHANNEL_ERROR BIT(1, UL) >>> + >>> +struct scmi_shared_mem { >>> + uint32_t reserved; >>> + uint32_t channel_status; >>> + uint32_t reserved1[2]; >>> + uint32_t flags; >>> + uint32_t length; >>> + uint32_t msg_header; >>> + uint8_t msg_payload[]; >>> +}; >> Should it be __packed? > The layout already matches the shared-memory format without padding: > Every fixed member is a uint32_t or an array of uint8_t, and the > mapping is page aligned. > > Therefore __packed would not change these offsets, but would reduce the > structure's alignment guarantee. I would prefer to keep the natural > alignment and add compile-time offset checks to document the transport > layout. Does that address your concern? Yes. [...] >>> diff --git a/xen/arch/arm/firmware/scmi-smc-multiagent.c b/xen/arch/arm/firmware/scmi-smc-multiagent.c >>> new file mode 100644 >>> index 0000000000..6c46f6cd36 >>> --- /dev/null >>> +++ b/xen/arch/arm/firmware/scmi-smc-multiagent.c >>> @@ -0,0 +1,792 @@ >>> +/* SPDX-License-Identifier: GPL-2.0-only */ >>> +/* >>> + * SCI SCMI multi-agent driver, using SMC/HVC shmem as transport. >>> + * >>> + * Oleksii Moisieiev <oleksii_moisieiev@epam.com> >>> + * Copyright (c) 2025 EPAM Systems >>> + */ >>> + >>> +#include <xen/acpi.h> >>> + >>> +#include <xen/device_tree.h> >>> +#include <xen/init.h> >>> +#include <xen/iocap.h> >>> +#include <xen/err.h> >>> +#include <xen/libfdt/libfdt.h> >>> +#include <xen/string.h> >>> +#include <xen/param.h> >>> +#include <xen/sched.h> >>> +#include <xen/vmap.h> >>> + >>> +#include <asm/firmware/sci.h> >>> +#include <asm/smccc.h> >>> + >>> +#include "scmi-proto.h" >>> +#include "scmi-shmem.h" >>> + >>> +struct scmi_channel { >>> + uint32_t agent_id; >> If all agent_ids large than UINT8_MAX are invalid, why do you need >> uint32_t here? > > According to the DEN0056F document: > > https://support.arm.com/documentation/den0056/f/?lang=en > > agent_id is uint32_t in SCMI messages. > Okay, so agent_id is 32-bit value. I can imagine a platform that have agent IDs 0x100, 0x200, 0x300 and so on. I believe, in this case your code will not work as intended, because you limit agent id to 8 bits in domain structure. [...] >>> +static int collect_agent_id(struct scmi_channel *agent_channel) >>> +{ >>> + int ret; >>> + scmi_msg_header_t hdr; >>> + struct scmi_msg_base_discover_agent_p2a da_rx; >>> + struct scmi_msg_base_discover_agent_a2p da_tx; >>> + >>> + ret = map_channel_memory(agent_channel); >>> + if ( ret ) >>> + return ret; >>> + >>> + hdr.id = SCMI_BASE_DISCOVER_AGENT; >>> + hdr.type = 0; >>> + hdr.protocol = SCMI_BASE_PROTOCOL; >>> + >>> + da_tx.agent_id = agent_channel->agent_id; >>> + >>> + ret = do_smc_xfer(agent_channel, &hdr, &da_tx, sizeof(da_tx), &da_rx, >>> + sizeof(da_rx)); >>> + if ( agent_channel->domain_id != DOMID_XEN ) >>> + unmap_channel_memory(agent_channel); >>> + if ( ret ) >> On error, Xen's channel will be mapped still. Is this intended behaviour? > It remains mapped when collect_agent_id() returns, but the existing > caller cleans it up. scmi_probe() marks the management channel DOMID_XEN > before discovery. If discovery fails, it jumps to error, calls > unmap_channel_memory(channel), then free_channel_list(). Thus the current > probe failure path does not leak the mapping. > > > On successful discovery the management mapping is deliberately retained > for subsequent permission and reset requests. Secondary discovery > mappings are always removed by collect_agent_id(), whether the transfer > succeeds or fails. Ah, okay. Thank you for clarification. [...] -- WBR, Volodymyr ^ permalink raw reply [flat|nested] 10+ messages in thread
* [PATCH v14 6/6] docs: arm: add SCI SCMI SMC multi-agent driver docs 2026-10-20 8:43 [PATCH v14 0/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent support Oleksii Moisieiev ` (3 preceding siblings ...) 2026-10-20 8:43 ` [PATCH v14 4/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent driver Oleksii Moisieiev @ 2026-10-20 8:43 ` Oleksii Moisieiev 2026-10-20 8:43 ` [PATCH v14 5/6] tools/xl/libxl: wire up SCMI SMC multi-agent configuration Oleksii Moisieiev 5 siblings, 0 replies; 10+ messages in thread From: Oleksii Moisieiev @ 2026-10-20 8:43 UTC (permalink / raw) To: xen-devel@lists.xenproject.org Cc: Andrew Cooper, Anthony PERARD, Bertrand Marquis, Jan Beulich, Juergen Gross, Julien Grall, Michal Orzel, Oleksii Moisieiev, Roger Pau Monné, Stefano Stabellini, Volodymyr Babchuk, Grygorii Strashko From: Grygorii Strashko <grygorii_strashko@epam.com> Add SCI SCMI SMC multi-agent driver documentation. It includes a detailed description of the SCMI multi-agent driver. This document explains the driver's functionality, configuration, and the compilation process. The Xen SCMI multi-agent driver is designed to provide SCMI access to system resources from different domains. Signed-off-by: Grygorii Strashko <grygorii_strashko@epam.com> Signed-off-by: Oleksii Moisieiev <oleksii_moisieiev@epam.com> --- Changes in v14: - describe secondary agents as xen,scmi-agent nodes under agents - update host, discovery and dom0less examples for the child-node format - clarify shared-memory placement and platform RAM reservations Changes in v10: - rephrase section about /firmware/scmi. Mentioned that this node is taken from Host DT and copied unmodified. - fix xen,reg address for secondary domains for Dom0less configuration Changes in v8: - update documentation to match the last DT format - fixed RST: "... code-block:: dts" -> ".. code-block:: dts" - update documentation with dom0less configuration example - update documentation with new param xen,dom0-sci-agent-id instead of the command line parameter Changes in v7: - update documentation in section of the xen_scmi configuration which is matched by "xen,sci" compatible instead of the direct path. Changes in v6: - remove all HVC mentions from the multi-agent doc - update sci-agent-id parameter description in the documentation - add missing Sign-of - minor fixes across the document Changes in v5: - rework multi-agent driver to leave Host Device-tree unmodified .../arm/firmware/arm-scmi.rst | 405 ++++++++++++++++++ 1 file changed, 405 insertions(+) diff --git a/docs/hypervisor-guide/arm/firmware/arm-scmi.rst b/docs/hypervisor-guide/arm/firmware/arm-scmi.rst index d9698f4e4b..25459d22dc 100644 --- a/docs/hypervisor-guide/arm/firmware/arm-scmi.rst +++ b/docs/hypervisor-guide/arm/firmware/arm-scmi.rst @@ -36,6 +36,8 @@ The below sections describe SCMI support options available for Xen. | [1] `Arm SCMI <https://developer.arm.com/documentation/den0056/latest/>`_ | [2] `System Control and Management Interface (SCMI) bindings <https://web.git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/firmware/arm,scmi.yaml>`_ +| [3] `Generic Domain Access Controllers bindings <https://web.git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/access-controllers/access-controllers.yaml>`_ + Simple SCMI over SMC calls forwarding driver (EL3) ------------------------------------------------------ @@ -189,3 +191,406 @@ except explicitly enabling SCMI with "arm_sci" xl.cfg option. -> xen,reg = <0x0 0x47ff0000 0x0 0x1000 0x0 0x22001000>; -> xen,force-assign-without-iommu; }; + +SCMI SMC multi-agent driver (EL3) +------------------------------------- + +The SCMI SMC multi-agent driver enables support for ARM EL3 Trusted Firmware-A (TF-A) which +provides SCMI interface with multi-agent support, as shown below. + +:: + + +-----------------------------------------+ + | | + | EL3 TF-A SCMI | + +-------+--+-------+--+-------+--+-------++ + |shmem1 | |shmem0 | |shmem2 | |shmemX | + +-----+-+ +---+---+ +--+----+ +---+---+ + smc-id1 | | | | + agent1 | | | | + +-----v--------+---------+-----------+----+ + | | | | | + | | | | | + +--------------+---------+-----------+----+ + smc-id0 | smc-id2| smc-idX| + agent0 | agent2 | agentX | + | | | + +----v---+ +--v-----+ +--v-----+ + | | | | | | + | Dom0 | | Dom1 | | DomX | + | | | | | | + | | | | | | + +--------+ +--------+ +--------+ + +The EL3 SCMI multi-agent firmware is expected to provide SCMI SMC shared-memory transport +for every Agent in the system. The SCMI Agent transport channel defined by pair: + +- smc-id: SMC function id used for Doorbell +- shmem: shared memory for messages transfer, **Xen page aligned**. + Shared memory is mapped with the following flags: MT_DEVICE_nGnRE and _PAGE_DEVICE, indicating that this + memory is mapped as device memory. + +The following SCMI Agents are expected to be defined by SCMI FW to enable SCMI multi-agent functionality +under Xen: + +- Xen management agent: trusted agents that accesses to the Base Protocol commands to configure + agent specific permissions +- OSPM VM agents: non-trusted agent, one for each Guest domain which is allowed direct HW access. + At least one OSPM VM agent has to be provided by FW if HW is handled only by Dom0 or Driver Domain. + +The EL3 SCMI FW is expected to implement following Base protocol messages: + +- BASE_DISCOVER_AGENT (required for Xen; required for secondary agents without ``reg``) +- BASE_RESET_AGENT_CONFIGURATION (optional) +- BASE_SET_DEVICE_PERMISSIONS (optional) + +The number of supported SCMI agents and their transport specifications are SCMI FW implementation +specific. + +Compiling with multi-agent support +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +To build with the SCMI SMC multi-agent driver support, enable Kconfig option: + +:: + + CONFIG_SCMI_SMC_MA + + +Driver functionality +^^^^^^^^^^^^^^^^^^^^ + +The SCI SCMI SMC multi-agent driver implements following functionality: + +- The driver is initialized from the Xen SCMI container ``xen_scmi_config`` + under ``/chosen/xen`` (for example ``/chosen/xen/xen_scmi_config/scmi``). + Only one SCMI interface is supported. The SCMI configuration must live under + the Xen SCMI container ``xen,sci`` beneath ``/chosen``. + The Xen SCMI mediator will bind only to the "arm,scmi-smc" node that is a child of + this "xen,sci" container; any other "arm,scmi-smc" nodes (for example under + "/firmware") are ignored to avoid stealing the host's SCMI OSPM instance. + +.. code-block:: dts + + scmi_shm_1: sram@47ff1000 { + compatible = "arm,scmi-shmem"; + reg = <0x0 0x47ff1000 0x0 0x1000>; + }; + scmi_xen: scmi { + compatible = "arm,scmi-smc"; + arm,smc-id = <0x82000003>; /* Xen management agent */ + #address-cells = < 1>; + #size-cells = < 0>; + #access-controller-cells = < 1>; + shmem = <&scmi_shm_1>; /* Xen management shared memory */ + }; + +.. note:: + This layout keeps the Host DT unchanged for Dom0 and baremetal Linux by + using func_id 0x82000002 / shmem 0x47ff0000 for Dom0, while Xen uses a + separate privileged channel func_id 0x82000003 / shmem 0x47ff1000. EL3 + firmware enforces permissions per agent_id, so there is no conflict between + Dom0 and Xen channels. + +- The driver reads secondary agents from an ``agents`` child of the + ``xen,sci`` container. This child has ``#address-cells = <1>`` and + ``#size-cells = <0>``. Each enabled ``xen,scmi-agent`` child specifies its + doorbell in ``arm,smc-id`` and a single shared-memory phandle in ``shmem``. + The optional single-cell ``reg`` property supplies the SCMI agent ID + (0..254). IDs must be unique and distinct from Xen's management agent. + At least one enabled secondary agent is required. + + The shared-memory nodes remain under ``xen_scmi_config``; the host OSPM + shared-memory node and ``/firmware/scmi`` remain in the platform DT. + A shared-memory node describes a channel location and does not reserve RAM. + If buffers occupy system RAM, the platform must reserve the entire + firmware-owned range in ``/reserved-memory`` before Xen boots. A single + reservation may cover multiple channels, including unused ones. SRAM + outside system RAM does not require a RAM reservation. The channel nodes + do not have to be moved into ``/reserved-memory``. + +.. code-block:: dts + + / { + #address-cells = <2>; + #size-cells = <2>; + + scmi_shm_0: sram@47ff0000 { + compatible = "arm,scmi-shmem"; + reg = <0x0 0x47ff0000 0x0 0x1000>; + }; + + firmware { + scmi: scmi { + compatible = "arm,scmi-smc"; + arm,smc-id = <0x82000002>; /* Host OSPM agent smc-id */ + #address-cells = <1>; + #size-cells = <0>; + shmem = <&scmi_shm_0>; /* Host OSPM agent shmem */ + }; + }; + + chosen { + #address-cells = <2>; + #size-cells = <2>; + ranges; + + xen { + #address-cells = <2>; + #size-cells = <2>; + ranges; + + xen_scmi_config { + compatible = "xen,sci"; + #address-cells = <2>; + #size-cells = <2>; + ranges; + xen,dom0-sci-agent-id = <0>; + + /* Xen management channel */ + scmi_shm_1: sram@47ff1000 { + compatible = "arm,scmi-shmem"; + reg = <0x0 0x47ff1000 0x0 0x1000>; + }; + + scmi_shm_2: sram@47ff2000 { + compatible = "arm,scmi-shmem"; + reg = <0x0 0x47ff2000 0x0 0x1000>; + }; + + scmi_shm_3: sram@47ff3000 { + compatible = "arm,scmi-shmem"; + reg = <0x0 0x47ff3000 0x0 0x1000>; + }; + + scmi_xen: scmi { + compatible = "arm,scmi-smc"; + arm,smc-id = <0x82000003>; /* Xen management smc-id */ + #address-cells = <1>; + #size-cells = <0>; + #access-controller-cells = <1>; + shmem = <&scmi_shm_1>; /* Xen management agent shmem */ + }; + + agents { + #address-cells = <1>; + #size-cells = <0>; + + agent@0 { + compatible = "xen,scmi-agent"; + reg = <0>; + arm,smc-id = <0x82000002>; + shmem = <&scmi_shm_0>; + }; + + agent@2 { + compatible = "xen,scmi-agent"; + reg = <2>; + arm,smc-id = <0x82000004>; + shmem = <&scmi_shm_2>; + }; + + agent@3 { + compatible = "xen,scmi-agent"; + reg = <3>; + arm,smc-id = <0x82000005>; + shmem = <&scmi_shm_3>; + }; + }; + }; + }; + }; + }; + +The ``xen,sci`` subtree is omitted from Dom0's DT. The host OSPM channel is +copied to Dom0 when SCMI is enabled for it. Xen uses its separate privileged +management channel to configure the other agents. + +To discover a secondary agent ID through ``BASE_DISCOVER_AGENT``, omit +``reg`` and use a node name without a unit address. For example, replace +only ``agent@2`` above with: + +.. code-block:: dts + + agent-domu { + compatible = "xen,scmi-agent"; + arm,smc-id = <0x82000004>; + shmem = <&scmi_shm_2>; + }; + +Providing ``reg`` skips mapping and probing that secondary channel during +initialization. A malformed ``reg`` is rejected. Xen always discovers its +own management agent ID. Thus ``BASE_DISCOVER_AGENT`` is mandatory for the +management channel and for secondary channels whose ``reg`` is omitted. +The former flat secondary-agent properties are replaced by these nodes. + +.. note:: + + Note that Xen is the only one entry in the system which need to know about SCMI multi-agent support. + +- The driver implements the SCI subsystem interface required for configuring and enabling SCMI + functionality for Dom0/hwdom and Guest domains. To enable SCMI functionality for guest domain + it has to be configured with unique supported SCMI Agent_id and use corresponding SCMI SMC + shared-memory transport ``[smc-id, shmem]`` defined for this SCMI Agent_id. + +- Once Xen domain is configured it can communicate with EL3 SCMI FW: + + - zero-copy, the guest domain puts/gets SCMI message in/from shmem; + - the guest triggers SMC exception with agent "smc-id" (doorbell); + - the Xen driver catches exception, do checks and synchronously forwards it to EL3 FW. + +- the Xen driver sends BASE_RESET_AGENT_CONFIGURATION message to Xen management agent channel on + domain destroy event. This allows to reset resources used by domain and so implement use-case + like domain reboot. + + +Configure SCMI for Dom0 +^^^^^^^^^^^^^^^^^^^^^^^ +Set the Dom0 SCMI agent ID in the device tree using the Xen SCMI container under ``/chosen``. +Add ``xen,dom0-sci-agent-id`` to the ``xen,sci`` node. If the property is absent, SCMI stays +disabled for Dom0 and the SCMI nodes are removed from Dom0 DT. + +.. code-block:: dts + + chosen { + xen { + ranges; + xen_scmi_config { + compatible = "xen,sci"; + xen,dom0-sci-agent-id = <0>; /* Dom0 agent ID */ + /* agents, shared memory and scmi_xen as shown above */ + }; + }; + }; + +The Host DT ``/firmware/scmi`` node is copied to the Dom0 DT unmodified. However, for Dom0 SCMI +configuration, Xen relies on the ``agents`` child and the ``xen,dom0-sci-agent-id`` +property from the ``xen,sci`` container under ``/chosen``. If the ``/firmware/scmi`` node is +missing or disabled, or if ``xen,dom0-sci-agent-id`` is not provided, the Dom0 SCMI agent will not +be configured. + +.. note:: + + The ``xen,dom0-sci-agent-id`` value must match the ``func_id`` and ``shmem`` pairing provided by + the EL3 firmware for Dom0 (for example in the ``/firmware/scmi`` node). + +Configure SCMI for for guest domain with toolstack +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +* In domain's xl.cfg file add **"arm_sci"** option as below + +:: + + arm_sci = "type=scmi_smc_multiagent,agent_id=2" + +* In domain's xl.cfg file enable access to the "arm,scmi-shmem" which should correspond + assigned "agent_id" for the domain, for example: + +:: + + iomem = [ + "47ff2,1@22001", + ] + +.. note:: It's up to the user to select guest IPA for mapping SCMI shared-memory. + +* Add SCMI nodes to the Driver domain partial device tree as in the below example. + The "arm,smc-id" should correspond assigned agent_id for the domain: + +.. code:: + + passthrough { + scmi_shm_0: sram@22001000 { + compatible = "arm,scmi-shmem"; + reg = <0x0 0x22001000 0x0 0x1000>; + }; + + firmware { + compatible = "simple-bus"; + scmi: scmi { + compatible = "arm,scmi-smc"; + arm,smc-id = <0x82000004>; <--- smc-id for agent_id=2 + shmem = <&scmi_shm_0>; + ... + } + } + } + +**Device specific access control** + +The XEN SCMI SMC multi-agent driver performs "access-controller" provider function in case +EL3 SCMI FW implements SCMI "4.2.1.1 Device specific access control" and provides the +BASE_SET_DEVICE_PERMISSIONS command to configure the devices that an agents have access to. +The Host DT SCMI node should have "#access-controller-cells=<1>" property and DT devices should +be bound to the SCMI node using Access Controllers bindings [3]. + +For example: + +.. code-block:: dts + + &i2c1 { + access-controllers = <&scmi 0>; + }; + +Use domain's xl.cfg file **"dtdev"** property to assign SCMI devices from toolstack to the guest: + +:: + + dtdev = [ + "/soc/i2c@e6508000", + ] + +.. note:: + + xl.cfg:"dtdev" need contain all nodes which are under SCMI management (not only those which are + behind IOMMU) and passed-through to the guest domain. + +Configure SCMI for predefined domains (dom0less) +^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + +* add "xen,sci_type" and "xen,sci-agent-id" properties for required DomU ("xen,domain") node + +:: + + xen,sci_type="scmi_smc_multiagent" + xen,sci-agent-id=2 + +* add scmi nodes to the Driver domain partial device tree the same way as above (toolstack case) and + enable access to the "arm,scmi-shmem" according to the dom0less documentation. For example: + +.. code-block:: dts + + scmi_shm_0: sram@22001000 { + compatible = "arm,scmi-shmem"; + reg = <0x00 0x22001000 0x00 0x1000>; + -> xen,reg = <0x0 0x47ff2000 0x0 0x1000 0x0 0x22001000>; + -> xen,force-assign-without-iommu; + }; + +* For SCMI device access control configure pass-through devices in the guest partial DT according to + the dom0less documentation and ensure that devices SCMI management has "xen,path" property set: + +Use the complete Xen SCMI configuration above. Remove +``xen,dom0-sci-agent-id`` if Dom0 does not use SCMI. The guest selects the +secondary channel described by ``agent@2``: + +.. code-block:: dts + + chosen { + xen,domain@1 { + compatible = "xen,domain"; + xen,sci_type = "scmi_smc_multiagent"; + xen,sci-agent-id = <2>; + /* other domain properties here */ + }; + }; + +.. code-block:: dts + + i2c@e6508000 { + ... + reg = <0x00 0xe6508000 0x00 0x1000>; + -> xen,path = "/soc/i2c@e6508000" + -> xen,reg = <0x0 0xe6508000 0x0 0x1000 0x0 0xe6508000>; + -> xen,force-assign-without-iommu; + }; -- 2.43.0 ^ permalink raw reply related [flat|nested] 10+ messages in thread
* [PATCH v14 5/6] tools/xl/libxl: wire up SCMI SMC multi-agent configuration 2026-10-20 8:43 [PATCH v14 0/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent support Oleksii Moisieiev ` (4 preceding siblings ...) 2026-10-20 8:43 ` [PATCH v14 6/6] docs: arm: add SCI SCMI SMC multi-agent driver docs Oleksii Moisieiev @ 2026-10-20 8:43 ` Oleksii Moisieiev 5 siblings, 0 replies; 10+ messages in thread From: Oleksii Moisieiev @ 2026-10-20 8:43 UTC (permalink / raw) To: xen-devel@lists.xenproject.org Cc: Andrew Cooper, Anthony PERARD, Bertrand Marquis, Jan Beulich, Juergen Gross, Julien Grall, Michal Orzel, Oleksii Moisieiev, Roger Pau Monné, Stefano Stabellini, Volodymyr Babchuk, Grygorii Strashko Plumb the SCMI SMC multi-agent type through the toolstack: - Extend libxl_arm_sci_type enumeration with scmi_smc_multiagent (value 2) - Add agent_id field to libxl_arm_sci structure for per-domain agent assignment - Update libxl_arm.c to translate libxl config to XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC_MA and pass agent_id to the hypervisor via xen_domctl_createdomain - Add xl.cfg parsing for arm_sci="type=scmi_smc_multiagent,agent_id=N" - Document the new xl.cfg options in xl.cfg.5.pod.in This completes the userspace side of multi-agent SCMI, allowing xl create and dom0less configurations to assign unique agent_id values to domains. Series-changes 12: - reject agent_id in xl.cfg when the ARM_SCI type is not scmi_smc_multiagent, instead of parsing it and discarding it - perform the check after the option loop, as type and agent_id may be given in either order Signed-off-by: Grygorii Strashko <grygorii_strashko@epam.com> Signed-off-by: Oleksii Moisieiev <oleksii_moisieiev@epam.com> --- (no changes since v11) Changes in v11: - Fix agent_id documentation: clarify it applies to SCMI SMC multi-agent support only, not plain SCMI SMC (reviewer feedback) - Remove "non-zero" from agent_id description to match accepted range [0..254] - Remove "UINT8_MAX (255) is treated as invalid" from user documentation as unnecessary implementation detail - Add LIBXL_HAVE_SCMI_SMC_MULTIAGENT feature macro in libxl.h to advertise the new scmi_smc_multiagent type and agent_id field - Add agent_id validation in libxl__arch_domain_build_info_setdefault() to reject invalid values at the libxl level, not only in xl Changes in v10: - Split hypervisor and toolstack changes into separate commits docs/man/xl.cfg.5.pod.in | 13 +++++++++++++ tools/include/libxl.h | 8 ++++++++ tools/libs/light/libxl_arm.c | 13 +++++++++++++ tools/libs/light/libxl_types.idl | 4 +++- tools/xl/xl_parse.c | 27 +++++++++++++++++++++++++++ 5 files changed, 64 insertions(+), 1 deletion(-) diff --git a/docs/man/xl.cfg.5.pod.in b/docs/man/xl.cfg.5.pod.in index b19c25c8d1..9d38ac0f7e 100644 --- a/docs/man/xl.cfg.5.pod.in +++ b/docs/man/xl.cfg.5.pod.in @@ -3184,8 +3184,21 @@ single SCMI OSPM agent support. Should be used together with B<scmi-smc-passthrough> Xen command line option. +=item B<scmi_smc_multiagent> + +Enables ARM SCMI SMC multi-agent support for the guest by enabling SCMI over +SMC calls forwarding from domain to the EL3 firmware (like Trusted Firmware-A) +with a multi SCMI OSPM agent support. The SCMI B<agent_id> should be +specified for the guest. + =back +=item B<agent_id=NUMBER> + +Specifies an ARM SCI agent id for the guest. This option is mandatory +if the SCMI SMC multi-agent support is enabled for the guest. The agent ids of +domains existing on a single host must be unique and in the range [0..254]. + =back =back diff --git a/tools/include/libxl.h b/tools/include/libxl.h index d294fb36bd..be25f23288 100644 --- a/tools/include/libxl.h +++ b/tools/include/libxl.h @@ -326,6 +326,14 @@ */ #define LIBXL_HAVE_BUILDINFO_ARCH_ARM_SCI 1 +/* + * LIBXL_HAVE_SCMI_SMC_MULTIAGENT indicates that the + * LIBXL_ARM_SCI_TYPE_SCMI_SMC_MULTIAGENT value is available in the + * libxl_arm_sci_type enumeration and the agent_id field is available + * in the libxl_arm_sci structure. + */ +#define LIBXL_HAVE_SCMI_SMC_MULTIAGENT 1 + /* * LIBXL_HAVE_SOFT_RESET indicates that libxl supports performing * 'soft reset' for domains and there is 'soft_reset' shutdown reason diff --git a/tools/libs/light/libxl_arm.c b/tools/libs/light/libxl_arm.c index aa8fa852fc..f92922abea 100644 --- a/tools/libs/light/libxl_arm.c +++ b/tools/libs/light/libxl_arm.c @@ -237,6 +237,10 @@ int libxl__arch_domain_prepare_config(libxl__gc *gc, case LIBXL_ARM_SCI_TYPE_SCMI_SMC: config->arch.arm_sci_type = XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC; break; + case LIBXL_ARM_SCI_TYPE_SCMI_SMC_MULTIAGENT: + config->arch.arm_sci_type = XEN_DOMCTL_CONFIG_ARM_SCI_SCMI_SMC_MA; + config->arch.arm_sci_agent_id = d_config->b_info.arch_arm.arm_sci.agent_id; + break; default: LOG(ERROR, "Unknown ARM_SCI type %d", d_config->b_info.arch_arm.arm_sci.type); @@ -1897,6 +1901,15 @@ int libxl__arch_domain_build_info_setdefault(libxl__gc *gc, } } + /* Sanitise ARM SCI agent_id parameter */ + if (b_info->arch_arm.arm_sci.type == LIBXL_ARM_SCI_TYPE_SCMI_SMC_MULTIAGENT && + b_info->arch_arm.arm_sci.agent_id >= UINT8_MAX) { + LOG(ERROR, + "Invalid ARM SCI agent_id: %u. Valid range is [0..254]", + b_info->arch_arm.arm_sci.agent_id); + return ERROR_FAIL; + } + if (b_info->type != LIBXL_DOMAIN_TYPE_PV) return 0; diff --git a/tools/libs/light/libxl_types.idl b/tools/libs/light/libxl_types.idl index d819955108..9b38c8f6de 100644 --- a/tools/libs/light/libxl_types.idl +++ b/tools/libs/light/libxl_types.idl @@ -555,11 +555,13 @@ libxl_sve_type = Enumeration("sve_type", [ libxl_arm_sci_type = Enumeration("arm_sci_type", [ (0, "none"), - (1, "scmi_smc") + (1, "scmi_smc"), + (2, "scmi_smc_multiagent") ], init_val = "LIBXL_ARM_SCI_TYPE_NONE") libxl_arm_sci = Struct("arm_sci", [ ("type", libxl_arm_sci_type), + ("agent_id", uint8) ]) libxl_rdm_reserve = Struct("rdm_reserve", [ diff --git a/tools/xl/xl_parse.c b/tools/xl/xl_parse.c index 01783731b5..6bb3325c98 100644 --- a/tools/xl/xl_parse.c +++ b/tools/xl/xl_parse.c @@ -1290,6 +1290,7 @@ static int parse_arm_sci_config(XLU_Config *cfg, libxl_arm_sci *arm_sci, const char *str) { int ret = 0; + int agent_id_set = 0; char *buf2, *ptr; char *oparg; @@ -1308,9 +1309,35 @@ static int parse_arm_sci_config(XLU_Config *cfg, libxl_arm_sci *arm_sci, } } + if (MATCH_OPTION("agent_id", ptr, oparg)) { + unsigned long val = parse_ulong(oparg); + + if ( val >= UINT8_MAX ) { + fprintf(stderr, "An invalid ARM_SCI agent_id specified (%lu). Valid range [0..254]\n", + val); + ret = ERROR_INVAL; + goto out; + } + arm_sci->agent_id = val; + agent_id_set = 1; + } + ptr = strtok(NULL, ","); } + /* + * agent_id only means something to the multi-agent driver, and + * libxl__arch_domain_prepare_config() only reads it for that type. + * Reject it elsewhere instead of silently dropping it. The check is done + * after the loop because the options may be given in any order. + */ + if (agent_id_set && + arm_sci->type != LIBXL_ARM_SCI_TYPE_SCMI_SMC_MULTIAGENT) { + fprintf(stderr, + "ARM_SCI agent_id is only supported with type=scmi_smc_multiagent\n"); + ret = ERROR_INVAL; + } + out: free(buf2); return ret; -- 2.43.0 ^ permalink raw reply related [flat|nested] 10+ messages in thread
end of thread, other threads:[~2026-10-29 18:56 UTC | newest] Thread overview: 10+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2026-10-20 8:43 [PATCH v14 0/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent support Oleksii Moisieiev 2026-10-20 8:43 ` [PATCH v14 1/6] xen/domctl: chain SCI handling before IOMMU in assign_device domctl Oleksii Moisieiev 2026-10-20 8:43 ` [PATCH v14 2/6] xen: arm: smccc: add INVALID_PARAMETER error code Oleksii Moisieiev 2026-10-20 8:43 ` [PATCH v14 3/6] lib/arm: Add I/O memory copy helpers Oleksii Moisieiev 2026-10-20 8:43 ` [PATCH v14 4/6] xen/arm: scmi: introduce SCI SCMI SMC multi-agent driver Oleksii Moisieiev 2026-10-23 1:56 ` Volodymyr Babchuk 2026-10-26 15:39 ` Oleksii Moisieiev 2026-10-29 18:55 ` Volodymyr Babchuk 2026-10-20 8:43 ` [PATCH v14 6/6] docs: arm: add SCI SCMI SMC multi-agent driver docs Oleksii Moisieiev 2026-10-20 8:43 ` [PATCH v14 5/6] tools/xl/libxl: wire up SCMI SMC multi-agent configuration Oleksii Moisieiev
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.