The previous delegating/undelegating sequence was incorrect as per the
specification DDI0615, "Architecture Reference Manual Supplement, The
Realm Management Extension (RME), for Armv9-A" Sections A1.1.1 and
A1.1.2
Off topic:
- cleaning the gpt_is_gpi_valid and gpt_check_pass_overlap
Change-Id: Idb64d0a2e6204f1708951137062847938ab5e0ac
Signed-off-by: Robert Wakim <robert.wakim@arm.com>
Introduced by FF-A v1.1 we must forward a call to FFA_VERSION
to the SPMC so that the ffa version of the caller can be stored
for later use. Since the return of FFA_VERSION is not wrapped in
a FF-A call we need to use a direct message request to do this
forwarding. For the spmd_handler in the SPMC to hand off to the
correct function we use w2 to specify a target framework function.
Therefore we must update PSCI CPU_OFF to do this as well.
Change-Id: Ibaa6832b66f1597b3d65aa8986034f0c5916016d
Signed-off-by: Daniel Boulby <daniel.boulby@arm.com>
Add tsp service to check the value of the PSTATE DIT bit is as
expected and toggle it's value. This is used to ensure that
the DIT bit is maintained during a switch from the Normal to
Secure worlds and back.
Change-Id: I4e8bdfa6530e5e75925c0079d4fa2795133c5105
Signed-off-by: Daniel Boulby <daniel.boulby@arm.com>
The original design prevented the primary CPU from doing a
warm boot sequence. This patch allows the primary to do warm
boot as well.
Signed-off-by: Mark Dykes <mark.dykes@arm.com>
Change-Id: I6baa50c3dff3051ff8b3e5a922d340634f651867
This patch enable/disable SVE/FPU for Realms depending
upon it's state in NS.
When this feature is enabled, traps to EL3 on SVE/FPU access from
Realms are disabled. However, RMM must ensure that the Realm <-> NS
SVE/FPU registers are not corrupted by each other and Realms do
not leak information to NS.
Change-Id: I0a27a055787976507017b72879ba6458f066624e
Signed-off-by: Subhasish Ghosh <subhasish.ghosh@arm.com>
This patch allocates the RMI and GTSI FIDs from the reserved
range in Standard Secure Service call range of SMCCC.
Signed-off-by: Subhasish Ghosh <subhasish.ghosh@arm.com>
Change-Id: I82e77778882194c2a78ca6340788d53bab7c3a50
The RMI command handling in RMMD did not preserve x4 to x7 when
returning to NS caller. Although this is allowed for SMCCCv1.0, this is
not correct as per v1.1. This fixes the same by differentiating the
onward and backward path during SMC handling.
This patch also fixes an issue with the backward path wherein the first
argument was being truncated to 32 bits.
Change-Id: Ibc85d574d5a2178a763975ddb32e456a12e7dc88
Signed-off-by: Soby Mathew <soby.mathew@arm.com>
As spm_mm cannot handle SVE/SME usage in NS world so its better to give
compilation error when ENABLE_SVE_FOR_NS=1 or ENABLE_SME_FOR_NS=1.
Signed-off-by: Manish Pandey <manish.pandey2@arm.com>
Change-Id: I69dbb272ca681bb020501342008eda20d4c0b096
This patch adds two new compile time options to enable SME in TF-A:
ENABLE_SME_FOR_NS and ENABLE_SME_FOR_SWD for use in non-secure and
secure worlds respectively. Setting ENABLE_SME_FOR_NS=1 will enable
SME for non-secure worlds and trap SME, SVE, and FPU/SIMD instructions
in secure context. Setting ENABLE_SME_FOR_SWD=1 will disable these
traps, but support for SME context management does not yet exist in
SPM so building with SPD=spmd will fail.
The existing ENABLE_SVE_FOR_NS and ENABLE_SVE_FOR_SWD options cannot
be used with SME as it is a superset of SVE and will enable SVE and
FPU/SIMD along with SME.
Signed-off-by: John Powell <john.powell@arm.com>
Change-Id: Iaaac9d22fe37b4a92315207891da848a8fd0ed73
sha 4ce3e99a3 introduced printf format specifiers for fixed width
types, which uses PRI*64 instead of "ll" for 64 bit values.
Signed-off-by: Manish Pandey <manish.pandey2@arm.com>
Change-Id: I30472411467061d58cc6ee22407ed3bad2552751
Use long instead of long long on aarch64 for 64_t stdint types.
Introduce inttypes.h to properly support printf format specifiers for
fixed width types for such change.
Change-Id: I0bca594687a996fde0a9702d7a383055b99f10a1
Signed-off-by: Scott Branden <scott.branden@broadcom.com>
Updated FFA_FEATURES according to FF-A v1.1 in SPMC can also be used
to retrieve feature information, and should now accept other arguments
than just FF-A call IDs.
Signed-off-by: J-Alves <joao.alves@arm.com>
Change-Id: I02cc24a31ab3092ec1ce6fed1a9649ffe7136782
Assert condition:
1 Register secure timer(ppi=29) for sdei nmi watchdog;
2 kernel panic and then kdump;
While kdump, kernel mask all cores sdei, secure timer trigger
and go to handle_masked_trigger() and assert here:
assert(se->affinity == my_mpidr);
As kernel register with flag=0, mpidr=0 and TF-A set flag to
SDEI_REGF_RM_PE but leave mpidr=0. So set mpidr to fix his
assert issue.
Signed-off-by: Ming Huang <huangming@linux.alibaba.com>
Change-Id: Ia9182f40bde94fb004b46e2a72b186eb0ef05166
Define a handler in the SPMD to route secure interrupts occurring while
the normal world runs. On a Group1 Secure interrupt (with a GICv3 or a
Group0 interrupt on GICv2), the normal world is pre-empted to EL3 and
redirected to the SPMD/SPMC for further handling.
Signed-off-by: Olivier Deprez <olivier.deprez@arm.com>
Signed-off-by: Madhukar Pappireddy <madhukar.pappireddy@arm.com>
Change-Id: I1350d74048c5549a2af8da0ba004c08512cc006a
SDEI specified event numbers in hexadecimal format.
Change event number format to hexadecimal to make
it easier for the reader to recognize the proper event.
Change-Id: Iac7a91d0910316e0ad54a8f09bc17209e8c6adf6
Signed-off-by: Vasyl Gomonovych <vgomonovych@marvell.com>
This patch updates and refactors the GPT library and fixes bugs.
- Support all combinations of PGS, PPS, and L0GPTSZ parameters.
- PPS and PGS are set at runtime, L0GPTSZ is read from GPCCR_EL3.
- Use compiler definitions to simplify code.
- Renaming functions to better suit intended uses.
- MMU enabled before GPT APIs called.
- Add comments to make function usage more clear in GPT library.
- Added _rme suffix to file names to differentiate better from the
GPT file system code.
- Renamed gpt_defs.h to gpt_rme_private.h to better separate private
and public code.
- Renamed gpt_core.c to gpt_rme.c to better conform to TF-A precedent.
Signed-off-by: John Powell <john.powell@arm.com>
Change-Id: I4cbb23b0f81e697baa9fb23ba458aa3f7d1ed919
The changes include:
- A new build option (ENABLE_RME) to enable FEAT_RME
- New image called RMM. RMM is R-EL2 firmware that manages Realms.
When building TF-A, a path to RMM image can be specified using
the "RMM" build flag. If RMM image is not provided, TRP is built
by default and used as RMM image.
- Support for RMM image in fiptool
Signed-off-by: Zelalem Aweke <zelalem.aweke@arm.com>
Change-Id: I017c23ef02e465a5198baafd665a60858ecd1b25
TRP is a small test payload that implements Realm Monitor
Management (RMM) functionalities. RMM runs in the Realm world
(R-EL2) and manages the execution of Realm VMs and their
interaction with the hypervisor in Normal world.
TRP is used to test the interface between RMM and Normal world
software, known as Realm Management Interface (RMI). Current
functions includes returning RMM version and transitioning
granules from Non-secure to Realm world and vice versa.
More information about RMM can be found at:
https://developer.arm.com/documentation/den0125/latest
Signed-off-by: Zelalem Aweke <zelalem.aweke@arm.com>
Change-Id: Ic7b9a1e1f3142ef6458d40150d0b4ba6bd723ea2
This patch introduces the RMM dispatcher into BL31. This
will be the mechanism that will enable communication to
take place between the Realm and non-secure world. Currently
gives the capability for granules to be
transitioned from non-secure type to realm and vice versa.
Signed-off-by: Zelalem Aweke <zelalem.aweke@arm.com>
Signed-off-by: Subhasish Ghosh <subhasish.ghosh@arm.com>
Change-Id: I1fdc99a4bdd42bc14911aa0c6954b131de309511
Defining SMC IDs for FF-A v1.1 notifications functionality, and adding
them to SPMD SMC handler, to ensure calls are forwarded to the SPMC.
Signed-off-by: J-Alves <joao.alves@arm.com>
Change-Id: Icc88aded0fd33507f7795e996bd4ff1c2fe679c8
This change refactors the SPMD to setup SPMC CPU contexts once and early
from spmd_spmc_init (single call to cm_setup_context rather than on each
and every warm boot).
Pass the core linear ID through a GP register as an implementation
defined behavior helping FF-A adoption to legacy TOSes (essentially
when secure virtualization is not used).
A first version of this change was originally submitted by Lukas [1].
Pasting below the original justification:
Our TEE, Kinibi, is used to receive the core linear ID in the x3
register of booting secondary cores.
This patch is necessary to bring up secondary cores with Kinibi as an
SPMC in SEL1.
In Kinibi, the TEE is mostly platform-independent and all platform-
specifics like topology is concentrated in TF-A of our customers.
That is why we don't have the MPIDR - linear ID mapping in Kinibi.
We need the correct linear ID to program the GICv2 target register,
for example in power management case.
It is not needed on GICv3/v4, because of using a fixed mapping from
MPIDR to ICDIPTR/GICD_ITARGETSRn register.
For debug and power management purpose, we also want a unified view to
linear id between Linux and the TEE.
E.g. to disable a core, to see what cores are printing a trace /
an event.
In the past, Kinibi had several other designs, but the complexity was
getting out of control:
* Platform-specific assembler macros in the kernel.
* A per-core SMC from Linux to tell the linear ID after the boot.
* With DynamiQ, it seems SIPs were playing with MPIDR register values,
reusing them between cores and changing them during boot.
[1] https://review.trustedfirmware.org/c/TF-A/trusted-firmware-a/+/10235
Signed-off-by: Olivier Deprez <olivier.deprez@arm.com>
Signed-off-by: Lukas Hanel <lukas.hanel@trustonic.com>
Change-Id: Ifa8fa208e9b8eb1642c80b5f7b54152dadafa75e
Consider a RAS scenario:
Enter EL3 by sync exception, then call spm_mm_sp_call() enter
EL0s to handle this error, then call sdei_dispatch_event() to
inform OS. Finally, return back to OS from sync exception flow.
Similar flow is sgi_ras_intr_handler() in sgi_ras.c.
The icc_pmr_el1 register will be change in above flow:
1 cm_el1_sysregs_context_save(NON_SECURE);
-> ehf_exited_normal_world();
##icc_pmr_el1: 0xf8 => 0x80
2 spm_mm_sp_call();
3 sdei_dispatch_event();
4 ehf_activate_priority(sdei_event_priority(map));
##icc_pmr_el1: 0x80 => 0x60
5 restore_and_resume_ns_context();
-> ehf_exited_normal_world();
##return due to has_valid_pri_activations(pe_data) == 1
6 ehf_deactivate_priority(sdei_event_priority(map));
##icc_pmr_el1: 0x60 => 0x80
The icc_pmr_el1 was rewrote from 0xf8 to 0x80. This issue will
result in OS hang when eret to OS from RAS flow.
Move ehf_activate_priority(sdei_event_priority(map)) after
restore_and_resume_ns_context() can fix this issue.
Signed-off-by: Ming Huang <huangming@linux.alibaba.com>
Change-Id: If01ec55cf0aabf1594dece1ad50d3ec3406cdabc
The SDEI specification now says that during an SDEI
event handler dispatch the SPSR should be set according
to the TakeException() pseudocode function defined in
the Arm Architecture Reference Manual. This patch sets
the SPSR according to the function given in
ARM DDI 0487F.c page J1-7635
Change-Id: Id2f8f2464fd69c701d81626162827e5c4449b658
Signed-off-by: Daniel Boulby <daniel.boulby@arm.com>
The SPMC at S-EL2 manages S-EL1 execution contexts for SPs. The
currently running SP vCPU state is always saved when the SPMC exits to
SPMD. A fresh vCPU context is always restored when the SPMC is entered
from the SPMD and a SP resumed. For performance optimization reasons
this permits omitting the saving/restoring of the S-EL1 context from
within the EL3 SPMD on entering/exiting the SPMC. The S-EL2 SPMC and
NS-EL1 context save/restore remain done in the SPMD.
Signed-off-by: Olivier Deprez <olivier.deprez@arm.com>
Change-Id: I66413ed5983913791ff5c9fc03c590ee65c6ccd7
Add SMC wrappers for handshaking the existence
and basic parameter validation for the SMCCC/PCI
API. The actual read/write/segment validation is
implemented by a given platform which will enable
the API by defining SMC_PCI_SUPPORT.
Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
Change-Id: I4485ad0fe6003cec6f5eedef688914d100513c21
The SMCCC, part 3 indicates that only the bottom
32-bits of a 32-bit SMC call are valid. The upper
bits must be zero. Lets enforce that so standard
service code can assume its been called that way.
Signed-off-by: Jeremy Linton <jeremy.linton@arm.com>
Change-Id: I1bac50fbdc3b6ddca5fe2d1d1f96166a65ac4eb4
Handle calls to the FFA_SPM_ID_GET interface. If FFA_SPM_ID_GET is
invoked from the non-secure physical FF-A instance, return the SPMC id
(defined in the SPMC manifest). If FFA_SPM_ID_GET is invoked from
the secure physical FF-A instance (e.g. the SPMC), return the SPMD id.
Change-Id: Id6d4e96b1da2510386d344e09c4553dba01227ec
Signed-off-by: Daniel Boulby <daniel.boulby@arm.com>
Standard Secure Services, complying to the SMCCC specification, are
discoverable: Any user can do the SMC call, and derive from the return
value (-1) if the service is implemented. Consequently we should not
*warn* if BL31 does not implement a service, as some services (TRNG, for
instance) might never be implemented for devices, as they are lacking
hardware.
Short of dropping the existing warning message altogether, change the
level to VERBOSE, which should prevent it actually being printed in
normal situations.
This removes the pointless TF-A messages on the console when booting
Linux, as modern kernels now call the SOCID and the TRNG service
unconditionally.
Change-Id: I08b0b02e0f46322ebe0b40b3991c3c9b5bed4f97
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
In the case of a SP pre-empted by a non-secure interrupt, the SPMC
returns to the SPMD through the FFA_INTERRUPT ABI. It is then forwarded
to the normal world driver hinting the SP has to be resumed after the
non-secure interrupt has been serviced.
Signed-off-by: Olivier Deprez <olivier.deprez@arm.com>
Change-Id: I51a694dddcb8ea30fa84e1f11d018bc2abec0a56
The current SPM_MM implementations expects the SP image addresses
as static macros. This means platforms wanting to use dynamically
allocated memory addresses are left out. This patch gets sp_boot_info
at the beginning of spm_sp_setup function and uses member variables
of sp_boot_info to setup the context. So member variables of
struct sp_boot_info and consequently the context can be initialized
by static macros or dynamiclly allocated memory address..
Change-Id: I1cb75190ab8026b845ae20a9c6cc416945b5d7b9
Signed-off-by: Mayur Gudmeti <mgudmeti@nvidia.com>
FF-A specification states that error codes should be typed int32_t.
SPMD's uses uint64_t for return values, which if assigned with a signed
type would have sign extension, and change the size of the return from
32-bit to 64-bit.
Signed-off-by: J-Alves <joao.alves@arm.com>
Change-Id: I288ab2ffec8330a2fe1f21df14e22c34bd83ced3
Typically, interrupts for a specific security state get handled in the
same security execption level if the execution is in the same security
state. For example, if a non-secure interrupt gets fired when CPU is
executing in NS-EL2 it gets handled in the non-secure world.
However, interrupts belonging to the opposite security state typically
demand a world(context) switch. This is inline with the security
principle which states a secure interrupt has to be handled in the
secure world. Hence, the TSPD in EL3 expects the context(handle) for a
secure interrupt to be non-secure and vice versa.
The function "tspd_sel1_interrupt_handler" is the handler registered
for S-EL1 interrupts by the TSPD. Based on the above assumption, it
provides an assertion to validate if the interrupt originated from
non-secure world and upon success arranges entry into the TSP at
'tsp_sel1_intr_entry' for handling the interrupt.
However, a race condition between non-secure and secure interrupts can
lead to a scenario where the above assumptions do not hold true and
further leading to following assert fail.
This patch fixes the bug which causes this assert fail:
ASSERT: services/spd/tspd/tspd_main.c:105
BACKTRACE: START: assert
0: EL3: 0x400c128
1: EL3: 0x400faf8
2: EL3: 0x40099a4
3: EL3: 0x4010d54
BACKTRACE: END: assert
Change-Id: I359d30fb5dbb1429a4a3c3fff37fdc64c07e9414
Signed-off-by: Madhukar Pappireddy <madhukar.pappireddy@arm.com>
Add a lock and spin lock/unlock calls when accessing the fields of the
SPMD PM structure.
Signed-off-by: Olivier Deprez <olivier.deprez@arm.com>
Change-Id: I9bab705564dc1ba003c29512b1f9be5f126fbb0d
Remove the former impdef SPMD service for SPMC entry point
registration. Replace with FFA_SECONDARY_EP_REGISTER ABI
providing a single entry point address into the SPMC for
primary and secondary cold boot.
Signed-off-by: Olivier Deprez <olivier.deprez@arm.com>
Change-Id: I067adeec25fc12cdae90c15a616903b4ac4d4d83
when updata routing of an SDEI event, if the registration flags
is SDEI_REGF_RM_PE, need to updata the affinity of shared event.
Signed-off-by: Tony Xie <tony.xie@rock-chips.com>
Change-Id: Ie5d7cc4199253f6af1c28b407f712caac3092d06
The code to check for the presence of the TRNG service relies on
toolchain garbage collection, which is not enabled with -O0.
Add #ifdef guards around the call to the TRNG service handler to
cover builds without optimisation as well.
Change-Id: I08ece2005ea1c8fa96afa13904a851dec6b24216
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Robert Wakim
Olivier Deprez
Federico Recanati
Daniel Boulby
Bipin Ravi
Mark Dykes
Alexei Fedorov
Subhasish Ghosh
Soby Mathew
Manish Pandey
johpow01
Scott Branden
J-Alves
Ming Huang
Vasyl Gomonovych
Zelalem Aweke
Daniel Boulby
Jeremy Linton
Andre Przywara
Mayur Gudmeti
Madhukar Pappireddy
Tony Xie
Max Shvetsov