This provides a non-noop implementation of spinlocks for JFFS2 in terms
of RTEMS mutexes. POSIX spinlocks are not appropriate for the types of
actions that occur during a held JFFS2 spinlock and can cause bad
dispatch level conditions.
Ensure that a single locking session is carried over OOB writes
including the OOB read that is sometimes required. This removes the
possibility of another write occurring between the read and write that
would make the write incorrect.
Explicitly test the 64-bit integer division and modulo operations. They
are essential for the timekeeping services. On most 32-bit targets,
they need a software implementation.
Update #3716.
Measure the runtime of 1000 no-operation instructions. This may help to
investigate runtime measurement variations which do not result from code
generation or source code changes.
The runtime measurement of a function which does nothing and just
returns gives an estimate of the measurement overhead.
Update #3716.
Fix a potential stack corruption in uniprocessor configurations during
start multitasking.
The system initialization uses the interrupt stack. A first level
interrupt shall never interrupt a context which uses the interrupt
stack. Such a use would lead to stack corruption and undefined system
behaviour. Unfortunately, in uniprocessor configurations this was the
case. Multiprocessing is started using _CPU_Context_restore(). The
caller of this function (_Thread_Start_multitasking()) uses the
interrupt stack. Later we have in cpukit/score/cpu/sparc/cpu_asm.S:
mov %g1, %psr ! restore status register and
! **** ENABLE TRAPS ****
ld [%o1 + G5_OFFSET], %g5 ! restore the global registers
ld [%o1 + G7_OFFSET], %g7
! Load thread specific ISR dispatch prevention flag
ld [%o1 + ISR_DISPATCH_DISABLE_STACK_OFFSET], %o2
! Store it to memory later to use the cycles
ldd [%o1 + L0_OFFSET], %l0 ! restore the local registers
ldd [%o1 + L2_OFFSET], %l2
ldd [%o1 + L4_OFFSET], %l4
ldd [%o1 + L6_OFFSET], %l6
! Now restore thread specific ISR dispatch prevention flag
st %o2, [%g6 + PER_CPU_ISR_DISPATCH_DISABLE]
ldd [%o1 + I0_OFFSET], %i0 ! restore the input registers
ldd [%o1 + I2_OFFSET], %i2
ldd [%o1 + I4_OFFSET], %i4
ldd [%o1 + I6_FP_OFFSET], %i6
ldd [%o1 + O6_SP_OFFSET], %o6 ! restore the output registers
Between the ENABLE TRAPS and the restore of the output registers, we
still use the stack of the caller and interrupts may be enabled. If an
interrupt happens in this code block, the interrupt stack is
concurrently used which may lead to a crash.
Fix this by adding a new function _SPARC_Start_multiprocessing() for
uniprocessor configurations. This function first sets the stack pointer
to use the stack of the heir thread.
Close#4955.
Casts from uint64_t to void* must go through a uintptr_t cast to avoid
warnings when building with the ILP32 ABI since this is otherwise an
implicit truncation to 32bits for a pointer.
This reworks the JFFS2 delayed work queue to avoid use of
on-chain/off-chain semantics since they vary in behavior under
RTEMS_DEBUG and are not guaranteed to be safe to use in SMP systems.
This adds all delayed work structs to the chain on FS init and does not
remove them until umount.
Updates #4956
When a buffer is modified by both hardware components such as DMA and by
software components, the buffer cache state must be kept in sync so that
data is not accidentally thrown away during future invalidations.
When polling hardware registers in high performance situations, don't
rely on usleep or other standard sleep functions since they will
necessarily rely on kernel ticks to be woken up. This can easily cause
an immense reduction in throughput.
Make some performance tests compatible to other RTEMS versions which do
not allow changing the priority of a task which owns a priority
inheritance mutex.
Update #3716.
The <rtems/test-info.h> header file is required for every RTEMS test
program. Move the RTEMS test printer support to a dedicated header file
<rtems/test-printer.h>. This removes an unnecessary dependency to the
RTEMS printer support in <rtems/test-info.h>.
Tests using the RTEMS Testing Framework no longer depend on the
<rtems/printer.h>.
Doing the enabled-by processing just for the ldflags and just for the
link custom commands is confusing. Use an option instead which is
intended to be used for such use cases.
Add _CPU_Get_TLS_thread_pointer() to get the thread pointer which is
used to get the address of thread-local storage objects associated with
a thread.
Update #4920.
Fix rtems_configuration_get_interrupt_stack_size() for some code models.
The _ISR_Stack_size symbol has an arbitrary absolute address and may not
be representable in the code model used by the compiler.
Update #4953.
Store symbols with an arbitrary absolute address such as _TLS_Size,
_TLS_Alignment, _TLS_Data_size, and _TLS_BSS_size in an object to avoid issues
with some code models.
Update #4953.
The internal JFFS2 locking does not guarantee that delayed writes will
not step on other reads and writes to the device. This adds locking to
prevent that in the JFFS2 NAND interworking layer.
The changes here ensure correct cache maintenance around DMA operations.
One cache flush was missing and two cache invalidations occurred before
the corresponding read that would make them necessary.
This moves delayed work to a temporary chain to prevent a locking
inversion between the delayed work lock and the alloc_sem lock. Delayed
work is now processed after the delayed work lock is released. Locking
order is any JFFS2 locks before the delayed work lock.
This adds `-u__extendsfdf2` to the `ldflags` for the dl07, dl08, and
dl09 tests to force the inclusion of `__extendsfdf2` in the base image.
This function is part of the GCC software floating point library and is
used in the tests to convert floats to doubles when calling `printf`.
This allows for conditionally setting 'ldflags' in scripts that use
`link_cc` and `link_cxx`. The immediate use case is allowing a linker
flag to be used only for MicroBlaze builds of certain tests.
- Add the section alignment to the size as the allocator may not
provide correctly aligned memory
- Only include symbols in the section when locating symbols. The
powerpc was incorrectly adding SDATA BSS symbols to the BSS offset
overrunning the section
Closes#4950
Kinsey Moore
Joel Sherrill
Sebastian Huber
Chris Johns
Alex White