These are guaranteed to work for bitfields that are
larger then 32 bits wide.
Change-Id: I39a641f08a255478fae583947bced762950d12ff
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
The app-facing interface for configuring interrupts was never
formally defined, instead it was defined separately for each arch
in their respective arch-specific header files. Occasionally these
would go out of sync.
Now there is a single irq.h header which defines this interface.
To avoid runtime overhead, these map to _arch_* implementations of
each that must be defined in headers pulled in by arch/cpu.h.
Change-Id: I69afbeff31fd07f981b5b291f3c427296b00a4ef
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Looking at the IDT in a debugger is confusing, add a pretty-printing
function.
Change-Id: Iacc5e204e5d11e3e875c75ddf6d2e2e80b230299
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
We have a new policy: users should not be able to configure
an interrupt with "forbidden" priority levels, and any priority
levels with special semantics will be activated by flags.
Change-Id: I757c19cfedcb1d0938eaf4da348ddafb71b3e001
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Having priority levels 0 and 1 reserved on x86 due to implementation
details on how the CPU uses the vector table is confusing to users,
and makes it unnecessarily difficult to share drivers between arches.
Now on x86, priority levels 0 and 1 are available. Semantically, all
priority levels have had 2 subtracted from them.
It is no longer necessary to specify a priority level when the
vector itself is specified. If an IDT entry has a specific vector
associated with it, any priority argument is simply ignored.
In gen_idt, some simplifications have been made:
- The printed representation of a generated entry now fits on one line
- Some checks being done in validate_priority() were redundant, as
generate_interrupt_vector_bitmap() also ensures that there are
sufficient free vectors within a priority level.
Change-Id: I26669d8ee0a53f48fbc2283490a8c42d8b1daf8e
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
It's not a function and requires all its arguments to be build-time
constants. Make this more obvious to the end user to ease confusion.
Change-Id: I64107cf4d9db9f0e853026ce78e477060570fe6f
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Most systems have far less than 256 IRQ lines available, so
save some bytes in ROM by making this a config option.
On systems with MVIC, omit the table entirely as the mapping
is fixed.
The build cmd_gen_idt is slightly easier to read and will fail
immediately if any of the commands in the sequence error out.
Change-Id: I411f114557591e5cd96b618e6f79f97e8bedadf0
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
This bitfield is only needed to find unused vectors in the IDT
for installing dynamic interrupts.
Change-Id: I34ecd330774a0e50f240b4396527682eded29627
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Add sys_exc_connect() (and its x86-compatible alias nanoCpuExcConnect())
that allows connecting an exception handler at runtime.
The current implementation is a bit of a bastard, to avoid disturbing
the current implementation of the exception handlers. Instead of hooking
_exc_wrapper() in all vectors and adapting the exception handlers, the
current exception handlers are still hooked directly in the vectors.
When an exception is hooked at runtime, _exc_wrapper() gets installed in
the vector and the real handler gets inserted in _sw_exc_table; this
means that the scheme only works with non-XIP kernels.
This should be enhanced so that _exc_wrapper() is hooked in all vectors,
and that current exception handlers (for faults mostly) are reworked to
be inserted in the _sw_exc_table and wrapped in _exc_wrapper().
Change-Id: Icaa14f4835b57873d2905b7fbcbb94eeb3b247d1
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
Rename the function and allow it to handle the 'type'
argument, which is ignored in this case.
Change-Id: I3d3493bea4511b2d026747505e7e52c5acc85012
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
The ESF was built using the 'alias' names of the GPRs (a1, lr, pc, etc)
rather than their 'real' name (rN).
Change-Id: I49cae5e94869a79a3165dc7f2347d8cec39dbf67
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
On ARM, GPRs are often known by two names. E.g. the stack pointer is
both 'sp' and 'r13', the first parameter to a function 'a1' and 'r0',
etc. This macro allows defining them in a data structure, e.g. in the
ESF, without having to create a union, use the correct type, etc: it is
less error-prone and makes for shorter code.
sys_define_gpr_with_alias(name1, name2)
Change-Id: Ie4a6caa1ac23f26be4f7f0e05e9265f2655062cc
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
The image will be linked at a different address and with different
ROM/RAM sizes to allow running a bootloder image that loads a payload
image. The addresses/sizes depend on if it is a XIP image or not
(CONFIG_XIP), and in the case of a XIP image, if it is a bootloader
image (CONFIG_IS_BOOTLOADER) or not.
In the case of a bootloader, it is given the full ROM and
CONFIG_BOOTLOADER_SRAM_SIZE kB of SRAM. When not a XIP image, it is given
the full SRAM minus CONFIG_BOOTLOADER_SRAM_SIZE kB, and is linked at the
start of SRAM.
Change-Id: Ibbb693c7bff022f313dac40f21c04a61f4bed115
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
Adds C++ support to the build system.
Change-Id: Ice1e57a13598e7a48b0bf3298fc318f4ce012ee6
Signed-off-by: Dmitriy Korovkin <dmitriy.korovkin@windriver.com>
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
Adds extern "C" { } blocks to header files so that they can be
safely used by C++ source files.
Change-Id: Ia4db0c36a5dac5d3de351184a297d2af0df64532
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
We can save a great deal of RAM this way, it only needs to be
in RAM if dynamic interrupts are in use.
At some point this config option broke, probably when static
interrupts were introduced into the system.
To induce build (instead of runtime) errors when irq_connect_dynamic()
is used without putting the table in RAM, the dynamic interrupt
functions are now conditionally compiled.
Change-Id: I4860508746fd375d189390163876c59b6c544c9a
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
This is needed for setting up an image that runs entirely from SRAM,
including its vector table. Ensure integrity of relocated vector table
by using serialization instructions when moving the vector table to
ensure it has been fully written before something makes use of it.
Change-Id: I00c600d557c87c75847f67fbc42f1c2c16157608
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
This option misunderstands how XIP works. The IDT is ALWAYS in ROM,
the question is whether crt0 will copy it into RAM or not. You can't
save ROM space in this way.
Change-Id: I58025e3d71ead35730d0a5026213299b4fcb5eb9
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
The interrupt API has been redesigned:
- irq_connect() for dynamic interrupts renamed to irq_connect_dynamic().
It will be used in situations where the new static irq_connect()
won't work, i.e. the value of arguments can't be computed at build time
- a new API for static interrupts replaces irq_connect(). it is used
exactly the same way as its dynamic counterpart. The old static irq
macros will be removed
- Separate stub assembly files are no longer needed as the stubs are now
generated inline with irq_connect()
ReST documentation updated for the changed API. Some detail about the
IDT in ROM added, and an oblique reference to the internal-only
_irq_handler_set() API removed; we don't talk about internal APIs in
the official documentation.
Change-Id: I280519993da0e0fe671eb537a876f67de33d3cd4
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
We are checking against the wrong option, which does not exist.
Change-Id: Ied24daa0930bc4629750ea90f3ac6dbc45e87fff
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
On SysV ABI, the NANO_ESF parameter is passed in via the stack.
For IAMCU, this is instead expected to be in EAX.
_ExcEnter is currently using EAX to stash the return address of
the calling stub while it does a stack switch. Change it to use ECX
for this purpose, and if we are running with IAMCU place the
parameter in EAX instead of pushing it.
The output of the fault handler has been cleaned up a bit and it
now also includes the code segment.
Change-Id: I466e3990a26a1a82dd486f3d8af5395eab60b049
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
This was never implemented on ARC/ARM and has been superseded
by irq_offload().
Some checks that were only done with CONFIG_LOAPIC_DEBUG fall
under the category of 'shouldn't ever happen' and have been
converted into assertions, instead of propagating return values
which are largely never checked.
Change-Id: I4eedca05bb7b384c4f3aa41a4f037f221f4a9cfe
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
If we are not doing any dynamic interrupts or exceptions, we
can put the IDT in ROM and save a considerable amount of RAM,
up to 2K if the IDT is the default size of 256 entries.
The _interrupt_vectors_allocated table can also be put in ROM
if we're not using any dynamic interrupts.
We introduce a new Kconfig option to force the IDT to be in RAM
for situations where no dynamic IRQs are used, but ROM footprint
needs to be conserved.
Change-Id: I38c9f1a8837b4db9f3dea1caa008374a26cbbf1d
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
The routines _int_latency_start() and _int_latency_stop() have been
replaced by macros that evaluate to nothing when the kernel config
option INT_LATENCY_BENCHMARK is not enabled thereby giving a performance
boost to the x86 versions of irq_lock() and and irq_unlock().
Change-Id: Iabfa7bf001f5b8396e7bcf5eebd6b1aa342bac46
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
Change terminology and use SoC instead of platform. An SoC provides
features and default configurations available with an SoC. A board
implements the SoC and adds more features and IP block specific to the
board to extend the SoC functionality such as sensors and debugging
features.
Change-Id: I15e8d78a6d4ecd5cfb3bc25ced9ba77e5ea1122f
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Add support for compilers conforming to the IAMCU calling convention
as documented by
https://github.com/hjl-tools/x86-psABI/wiki/iamcu-psABI-0.7.pdf
Change-Id: I6fd9d5bede0538b2049772e3850a5940c5dd911e
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Add null definitions for the interrupt latency measurement API so we
can remove compile fences in C code.
Change-Id: If86eedf79afcb49002108814dd4fb864956eb667
Signed-off-by: Dirk Brandewie <dirk.j.brandewie@intel.com>
Removed old style file description and documnetation and apply
doxygen synatx.
Change-Id: I3ac9f06d4f574bf3c79c6f6044cec3a7e2f6e4c8
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Removes the 'priority' parameter from the IRQ_CONFIG macro.
This parameter was not used anymore in any architecture.
The priority is handled in the IRQ_CONNECT macro.
The documentation is updated as well.
Change-Id: I24a293c5e41bd729d5e759113e0c4a8a6a61e0dd
Signed-off-by: Juan Manuel Cruz <juan.m.cruz.alcaraz@linux.intel.com>
Adds static irq support for the Quark SE platform for the ARC core.
New linker sections and sw isr table initialization is needed to support
static IRQ.
Change-Id: I82af98a189f5a156e7f1018f3ecdbfa73ad3e6ef
Signed-off-by: Juan Manuel Cruz <juan.m.cruz.alcaraz@linux.intel.com>
Adds support for static IRQ handler initialization.
Currently, IRQ_CONNECT and IRQ_CONFIG macros are emulating static
behavior through dynamic initialization.
This commit updates the macros to get real static initialization.
IRQ handlers must be assigned at build time.
Change-Id: Ia07fb25a5e4dae489f84ffcedb28007ee18a3b82
Signed-off-by: Juan Manuel Cruz <juan.m.cruz.alcaraz@linux.intel.com>
IRQ_CONNECT_STATIC takes 6 arguments on other
architectures, but the ARC one had only 5.
Change-Id: I257e8db12582ee2d6f93bba63af9aa597197a53d
Signed-off-by: Juan Manuel Cruz <juan.m.cruz.alcaraz@linux.intel.com>
With MVIC these can't be arbitrarily assigned and the vector must be
<irq num> + 0x20.
The correct number of vectors is now set for footprint-min on D2000.
Change-Id: Ibf59921dbc438c7465b7050dd74d0badc9a91fc3
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Uses the "trap_s" exception to simulate entry into IRQ context;
offloaded functions run on the FIRQ stack.
Change-Id: I310ce42b45aca5dabd1d27e486645d23fa0b118f
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
This way, it does not fall in the middle of a group, like the RAM group
and as a side-effect potentially move the dot (current address pointer).
Change-Id: Iefbff8bbeadfc740dee61154d7db99b7b7aad6d6
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
The linker scripts for the quark_se_ss and generic_arc platforms are the
exact same, so extract the contents in an includable file.
Change-Id: I2cb90a6f819b12db77880228e41ff14c9755d59a
Signed-off-by: Benjamin Walsh <benjamin.walsh@windriver.com>
Initialization level can be one of five predefined.
Init priority is numeric from 1 to 99. If init level or priority
is defined wrong, linker prints out the message and stops.
Change-Id: I165a32ffb668cda983fd48eb2aa7b94998e31a18
Signed-off-by: Dmitriy Korovkin <dmitriy.korovkin@windriver.com>
Do not depend on environment variables and use a kconfig variable
for defining the architecture.
In addition, remove the X86_32 variable, it just duplicates X86 for
not good reason, at least until start supporting MCUs with 64bit.
Change-Id: Ia001db81ed007e6a43f34506fed9be1345b88a4b
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
These should now work for drivers written for other arches.
Still a hack to do all the IRQ setup at runtime.
Change-Id: I9717f74abef3b9934f9a1c0acbd76d960ed7a3cb
Signed-off-by: Andrew Boie <andrew.p.boie@intel.com>
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
The new flags parameter needs some documentation.
Change-Id: I24dc9df62323957bb4b294adf27487df3f76ea01
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Was exposed when building with clang. No need for a second const.
Change-Id: Ie97f6a4756aff62ce969e3eb786593f2fc175a56
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
Flags allow passing IRQ triggering option for x86 architecture.
Each platform defines flags for a particular device and then
device driver uses them when registers the interrupt handler.
The change in API means that device drivers and sample
applications need to use the new API.
IRQ triggering configuration is now handled by device drivers
by using flags passed to interrupt registering API:
IRQ_CONNECT_STATIC() or irq_connect()
Change-Id: Ibc4312ea2b4032a2efc5b913c6389f780a2a11d1
Signed-off-by: Dmitriy Korovkin <dmitriy.korovkin@windriver.com>
I/O ports are not memory and thus such asm instruction cannot follow
such constraint. Plus, usual BT* instruction can be used on normal
registers.
Change-Id: Ie3aad668173962a0a90e7cb11231c7843836d412
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Let the compiler decide about the registers to use, depending on which
functions those are called from (as they might be already in use or
not).
Change-Id: I00afa0f82c740c8ea70133d85ab67e9cb117187d
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Remove prototype support for C++ constructors, since it is not well
designed. Device drivers (or other application code) that requires an
automatic initialization capability should use the device initialization
macros instead.
Note: Support for C++ constructors may be re-introduced at a later date.
However, a number of issues need to be settled, such as when the
constructors are invoked and what context they run in. (Running them
during nanokernel initialization, as was previously done, is probably
not the right approach.)
Change-Id: If6d27ac16b485cb39d5ec34084e9d0f1991074f4
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
Since commit 3d7b21b69e the 'priority' parameter is used by the macro
IRQ_CONNECT_STATIC. This patch fixes the macro documentation which
says it is ignored.
Change-Id: I56eec49466a33441fa1822af78956ebb7f5c24ce
Signed-off-by: Andre Guedes <andre.guedes@intel.com>
As the system always operates in ring 0, neither the SS nor ESP registers
are pushed onto the stack when an exception or an interrupt occurs.
However, as the ESP field is still relevant to debugging fatal errors, a
place has been carved for it in the NANO_ESF.
Change-Id: Ibb2578c69fa6365fd6e9dbf7b51f461063dadc68
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
As page fault exceptions can not occur in the system as it is currently
designed, there is no need to track the CR2 register as part of the
exception stack frame.
Change-Id: I75d7a74c5d2c6efcc0e9141d2662861bc2052629
Signed-off-by: Peter Mitsis <peter.mitsis@windriver.com>
