The current code is incredibly resilient to updates to the spec and
has worked quite well so far. However, recent implementations expose a
weakness in that this is rather slow. A large part of it is written in
assembly, making it opaque to the compiler for optimisations. The
future proofness requires reading registers that are effectively
`volatile`, making it even harder for the compiler, as well as adding
lots of implicit barriers, making it hard for the microarchitecutre to
optimise as well.
We can make a few assumptions, checked by a few well placed asserts, and
remove a lot of this burden. For a start, at the moment there are 4
group 0 counters with static assignments. Contexting them is a trivial
affair that doesn't need a loop. Similarly, there can only be up to 16
group 1 counters. Contexting them is a bit harder, but we can do with a
single branch with a falling through switch. If/when both of these
change, we have a pair of asserts and the feature detection mechanism to
guard us against pretending that we support something we don't.
We can drop contexting of the offset registers. They are fully
accessible by EL2 and as such are its responsibility to preserve on
powerdown.
Another small thing we can do, is pass the core_pos into the hook.
The caller already knows which core we're running on, we don't need to
call this non-trivial function again.
Finally, knowing this, we don't really need the auxiliary AMUs to be
described by the device tree. Linux doesn't care at the moment, and any
information we need for EL3 can be neatly placed in a simple array.
All of this, combined with lifting the actual saving out of assembly,
reduces the instructions to save the context from 180 to 40, including a
lot fewer branches. The code is also much shorter and easier to read.
Also propagate to aarch32 so that the two don't diverge too much.
Change-Id: Ib62e6e9ba5be7fb9fb8965c8eee148d5598a5361
Signed-off-by: Boyan Karatotev <boyan.karatotev@arm.com>
Currently, EL3 context registers are duplicated per-world per-cpu.
Some registers have the same value across all CPUs, so this patch
moves these registers out into a per-world context to reduce
memory usage.
Change-Id: I91294e3d5f4af21a58c23599af2bdbd2a747c54a
Signed-off-by: Elizabeth Ho <elizabeth.ho@arm.com>
Signed-off-by: Jayanth Dodderi Chidanand <jayanthdodderi.chidanand@arm.com>
Align entire TF-A to use Arm in copyright header.
Change-Id: Ief9992169efdab61d0da6bd8c5180de7a4bc2244
Signed-off-by: Govindraj Raja <govindraj.raja@arm.com>
Combining the EL2 and EL3 enablement code necessitates that it must be
called at el3_exit, which is the only place with enough context to make
the decision of what needs to be set.
Decouple them to allow them to be called from elsewhere. Also take
some time to clarify and simplify AMU code.
The sanity check in the context_restore() is now wrong, as the cpu may
turn off on suspend, thus resetting the value of the counter enables.
Remove it.
Finally, this completes the migration to cm_manage_extensions_el3() and
manage_extensions_nonsecure() so manage_extensions_nonsecure_mixed() is
being removed.
Signed-off-by: Boyan Karatotev <boyan.karatotev@arm.com>
Change-Id: I66399132364c32be66017506bb54cbadd8485577
The AMU extension code was using its own feature detection routines.
Replace them with the generic CPU feature handlers (defined in
arch_features.h), which get updated to cover the v1p1 variant as well.
Change-Id: I8540f1e745d7b02a25a6c6cdf2a39d6f5e21f2aa
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
This change makes AMU auxiliary counters configurable on a per-core
basis, controlled by `ENABLE_AMU_AUXILIARY_COUNTERS`.
Auxiliary counters can be described via the `HW_CONFIG` device tree if
the `ENABLE_AMU_FCONF` build option is enabled, or the platform must
otherwise implement the `plat_amu_topology` function.
A new phandle property for `cpu` nodes (`amu`) has been introduced to
the `HW_CONFIG` specification to allow CPUs to describe the view of
their own AMU:
```
cpu0: cpu@0 {
...
amu = <&cpu0_amu>;
};
```
Multiple cores may share an `amu` handle if they implement the
same set of auxiliary counters.
AMU counters are described for one or more AMUs through the use of a new
`amus` node:
```
amus {
cpu0_amu: amu-0 {
#address-cells = <1>;
#size-cells = <0>;
counter@0 {
reg = <0>;
enable-at-el3;
};
counter@n {
reg = <n>;
...
};
};
};
```
This structure describes the **auxiliary** (group 1) AMU counters.
Architected counters have architecturally-defined behaviour, and as
such do not require DTB entries.
These `counter` nodes support two properties:
- The `reg` property represents the counter register index.
- The presence of the `enable-at-el3` property determines whether
the firmware should enable the counter prior to exiting EL3.
Change-Id: Ie43aee010518c5725a3b338a4899b0857caf4c28
Signed-off-by: Chris Kay <chris.kay@arm.com>
This change reduces the exposed surface area of the AMU API in order to
simplify the refactoring work in following patches. The functions and
definitions privatized by this change are not used by other parts of the
code-base today.
BREAKING CHANGE: The public AMU API has been reduced to enablement only
to facilitate refactoring work. These APIs were not previously used.
Change-Id: Ibf6174fb5b3949de3c4ba6847cce47d82a6bd08c
Signed-off-by: Chris Kay <chris.kay@arm.com>
With the introduction of MPMM, the auxiliary AMU counter logic requires
refactoring to move away from a single platform-defined group 1 counter
mask in order to support microarchitectural (per-core) group 1 counters.
BREAKING CHANGE: The `PLAT_AMU_GROUP1_COUNTERS_MASK` platform definition
has been removed. Platforms should specify per-core AMU counter masks
via FCONF or a platform-specific mechanism going forward.
Change-Id: I1e852797c7954f92409222b066a1ae57bc72bb05
Signed-off-by: Chris Kay <chris.kay@arm.com>
If SVE are enabled for both Non-secure and Secure world along with AMU
extension, then it causes the TAM_BIT in CPTR_EL3 to be set upon exit
from bl31. This restricts access to the AMU register set in normal
world. This fix maintains consistency in both TAM_BIT and CPTR_EZ_BIT
by saving and restoring CPTR_EL3 register from EL3 context.
Signed-off-by: Arunachalam Ganapathy <arunachalam.ganapathy@arm.com>
Change-Id: Id76ce1d27ee48bed65eb32392036377716aff087
ARMv8.6 adds virtual offset registers to support virtualization of the
event counters in EL1 and EL0. This patch enables support for this
feature in EL3 firmware.
Signed-off-by: John Powell <john.powell@arm.com>
Change-Id: I7ee1f3d9f554930bf5ef6f3d492e932e6d95b217
This patch fixes the bug when AMUv1 group1 counters was
always assumed being implemented without checking for its
presence which was causing exception otherwise.
The AMU extension code was also modified as listed below:
- Added detection of AMUv1 for ARMv8.6
- 'PLAT_AMU_GROUP1_NR_COUNTERS' build option is removed and
number of group1 counters 'AMU_GROUP1_NR_COUNTERS' is now
calculated based on 'AMU_GROUP1_COUNTERS_MASK' value
- Added bit fields definitions and access functions for
AMCFGR_EL0/AMCFGR and AMCGCR_EL0/AMCGCR registers
- Unification of amu.c Aarch64 and Aarch32 source files
- Bug fixes and TF-A coding style compliant changes.
Change-Id: I14e407be62c3026ebc674ec7045e240ccb71e1fb
Signed-off-by: Alexei Fedorov <Alexei.Fedorov@arm.com>
Enforce full include path for includes. Deprecate old paths.
The following folders inside include/lib have been left unchanged:
- include/lib/cpus/${ARCH}
- include/lib/el3_runtime/${ARCH}
The reason for this change is that having a global namespace for
includes isn't a good idea. It defeats one of the advantages of having
folders and it introduces problems that are sometimes subtle (because
you may not know the header you are actually including if there are two
of them).
For example, this patch had to be created because two headers were
called the same way: e0ea0928d5 ("Fix gpio includes of mt8173 platform
to avoid collision."). More recently, this patch has had similar
problems: 46f9b2c3a2 ("drivers: add tzc380 support").
This problem was introduced in commit 4ecca33988 ("Move include and
source files to logical locations"). At that time, there weren't too
many headers so it wasn't a real issue. However, time has shown that
this creates problems.
Platforms that want to preserve the way they include headers may add the
removed paths to PLAT_INCLUDES, but this is discouraged.
Change-Id: I39dc53ed98f9e297a5966e723d1936d6ccf2fc8f
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
The codebase was using non-standard headers. It is needed to replace
them by the correct ones so that we can use the new libc headers.
Change-Id: I530f71d9510cb036e69fe79823c8230afe890b9d
Acked-by: Sumit Garg <sumit.garg@linaro.org>
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
Add amu_context_save() and amu_context_restore() functions for aarch32
Change-Id: I4df83d447adeaa9d9f203e16dc5a919ffc04d87a
Signed-off-by: Joel Hutton <joel.hutton@arm.com>
Add some AMU helper functions to allow configuring, reading and
writing of the Group 0 and Group 1 counters. Documentation for these
helpers will come in a separate patch.
Change-Id: I656e070d2dae830c22414f694aa655341d4e2c40
Signed-off-by: Dimitris Papastamos <dimitris.papastamos@arm.com>
A new platform macro `PLAT_AMU_GROUP1_COUNTERS_MASK` controls which
group 1 counters should be enabled. The maximum number of group 1
counters supported by AMUv1 is 16 so the mask can be at most 0xffff.
If the platform does not define this mask, no group 1 counters are
enabled.
A related platform macro `PLAT_AMU_GROUP1_NR_COUNTERS` is used by
generic code to allocate an array to save and restore the counters on
CPU suspend.
Change-Id: I6d135badf4846292de931a43bb563077f42bb47b
Signed-off-by: Dimitris Papastamos <dimitris.papastamos@arm.com>
The `ENABLE_AMU` build option can be used to enable the
architecturally defined AMU counters. At present, there is no support
for the auxiliary counter group.
Change-Id: I7ea0c0a00327f463199d1b0a481f01dadb09d312
Signed-off-by: Dimitris Papastamos <dimitris.papastamos@arm.com>
