Searching for duplicate names scales O(n^2) with the number of names
added to a fdt, which can cause a noticable slowdown with larger device
trees and very slow CPU cores.
Add FDT_CREATE_FLAG_NO_NAME_DEDUP that allow the caller to trade fdt size
for speed in the creation process.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Message-Id: <20190509094122.834-4-npiggin@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
There is a need to be able to specify some options when building an FDT
with the SW interface. This can be accomplished with minimal changes by
storing intermediate data in the fdt header itself, in fields that are
not otherwise needed during the creation process and can be set by
fdt_finish().
The fdt.magic field is already used exactly this way, as a state to
check with callers that the FDT has been created but not yet finished.
fdt.version and fdt.last_comp_version are used to make room for more
intermediate state. These are adjacent and unused during the building
process. last_comp_version is not yet used for intermediate state, but
it is zeroed and treated as used, so as to allow future growth easily.
A new interface, fdt_create_with_flags() is added, which takes 32-bit
flag value to control creation.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Message-Id: <20190509094122.834-3-npiggin@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
If fdt_add_property or fdt_property_placeholder fail after allocating
a string for the name, they return without freeing that string. This
does not change the structure of the tree, but in very specific cases
it could lead to undesirable space consumption.
Fix this by rolling back the string allocation in this situation.
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Message-Id: <20190509094122.834-2-npiggin@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
It's now a trivial wrapper around fdt_find_max_phandle() so we might as
well inline it. We also remove it from the versioning linker script.
Theoretically, that's a breaking ABI change except that we haven't yet
released a version with it exposed in the shared object, so we can get
away with it.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The new fdt_generate_phandle() function can be used to generate a new,
unused phandle given a specific device tree blob. The implementation is
somewhat naive in that it simply walks the entire device tree to find
the highest phandle value and then returns a phandle value one higher
than that. A more clever implementation might try to find holes in the
current set of phandle values and fill them. But this implementation is
relatively simple and works reliably.
Also add a test that validates that phandles generated by this new API
are indeed unique.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Message-Id: <20190326153302.17109-3-thierry.reding@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The fdt_get_max_phandle() function has some shortcomings. On one hand
it returns just a uint32_t which means to check for the "negative"
error code a caller has to explicitly check against the error code
(uint32_t)-1. In addition, the -1 is the only error code that can be
returned, so a caller cannot tell the difference between the various
failures.
Fix this by adding a new fdt_find_max_phandle() function that returns an
error code on failure and 0 on success, just like other APIs, and stores
the maximum phandle value in an output argument on success.
This also refactors fdt_get_max_phandle() to use the new function. Add a
note pointing out that the new fdt_find_max_phandle() function should be
preferred over fdt_get_max_phandle().
Signed-off-by: Thierry Reding <treding@nvidia.com>
Message-Id: <20190326153302.17109-1-thierry.reding@gmail.com>
[dwg: Reword for some inaccuracies in the commit message]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
This function will append an address range property using parent node's
"#address-cells" and "#size-cells" properties.
It will be used in implementing kdump with kexec_file_load system call
at linux kernel for arm64 once it is merged into kernel tree.
Signed-off-by: AKASHI Takahiro <takahiro.akashi@linaro.org>
Message-Id: <20190327061552.17170-2-takahiro.akashi@linaro.org>
[dwg: Correct a SEGV error in the testcase]
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The new fdt_generate_phandle() function can be used to generate a new,
unused phandle given a specific device tree blob. The implementation is
somewhat naive in that it simply walks the entire device tree to find
the highest phandle value and then returns a phandle value one higher
than that. A more clever implementation might try to find holes in the
current set of phandle values and fill them. But this implementation is
relatively simple and works reliably.
Also add a test that validates that phandles generated by this new API
are indeed unique.
Signed-off-by: Thierry Reding <treding@nvidia.com>
Message-Id: <20190320151003.28941-1-thierry.reding@gmail.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
The actual error is FDT_ERR_NOTFOUND, not FDT_ERR_NOT_FOUND.
Fixes: d29126c90acb0d70 ("libfdt: Add iterator over properties")
Fixes: 902d0f0953d0074b ("libfdt: Add a subnodes iterator macro")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
6dcb8ba4 "libfdt: Add helpers for accessing unaligned words" introduced
the fdt32_ld() and fdt64_ld() helpers for loading values from the FDT blob
which might not be naturally aligned. This matters for ARM, where
attempting a plain unaligned load will often cause an exception.
However, it seems the memcpy() we used here was surprisingly expensive,
making libfdt nearly 6x slower on at least some ARM platforms.
This patch takes an alternative approach, using a bunch of 1-byte loads
and shifts to implement the helpers.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
According to the device tree specification, the default value for
#size-cells is 1, but fdt_size_cells() was returning 2 if this property
was not present.
This patch also makes fdt_address_cells() and fdt_size_cells() conform
to the behaviour documented in libfdt.h. The defaults are only returned
if fdt_getprop() returns -FDT_ERR_NOTFOUND, otherwise the actual error
is returned.
Signed-off-by: John Clarke <johnc@kirriwa.net>
In uniprocessor and SMP configurations, the context switch extensions
were called during _Thread_Do_dispatch():
void _Thread_Do_dispatch( Per_CPU_Control *cpu_self, ISR_Level level )
{
Thread_Control *executing;
executing = cpu_self->executing;
...
do {
Thread_Control *heir;
heir = _Thread_Get_heir_and_make_it_executing( cpu_self );
...
_User_extensions_Thread_switch( executing, heir );
...
_Context_Switch( &executing->Registers, &heir->Registers );
...
} while ( cpu_self->dispatch_necessary );
...
}
In uniprocessor configurations, this is fine and the context switch
extensions are called for all thread switches except the very first
thread switch to the initialization thread. However, in SMP
configurations, the context switch may be invalidated and updated in the
low-level _Context_Switch() routine. See:
https://docs.rtems.org/branches/master/c-user/symmetric_multiprocessing_services.html#thread-dispatch-details
In case such an update happens, a thread will execute on the processor
which was not seen in the previous call of the context switch
extensions. This can confuse for example event record consumers which
use events generated by a context switch extension.
Fixing this is not straight forward. The context switch extensions call
must move after the low-level context switch. The problem here is that
we may end up in _Thread_Handler(). Adding the context switch
extensions call to _Thread_Handler() covers now also the thread switch
to the initialization thread. We also have to save the last executing
thread (ancestor) of the processor. Registers or the stack cannot be
used for this purpose. We have to add it to the per-processor
information. Existing extensions may be affected, since now context
switch extensions use the stack of the heir thread. The stack checker
is affected by this.
Calling the thread switch extensions in the low-level context switch is
difficult since at this point an intermediate stack is used which is
only large enough to enable servicing of interrupts.
Update #3885.
Remove all comments and copyrightable content from the moved content.
Use BSD-2-Clause license for new file.
Change licence of <rtems/confdefs.h> to BSD-2-Clause according to file
history.
Update #3053.
Close#3875.
Remove all comments and copyrightable content from the moved content.
Use BSD-2-Clause license for new file.
Move the default configuration to library files so that application
configurations do not include the definitions for the default case.
Update #3053.
Update #3875.
Remove comments and copyrightable content from the moved content. Use
BSD-2-Clause for new file according to file history of
<rtems/confdefs.h>.
Update #3053.
Update #3875.
Unify handling of obsolete configuration options. Remove comments and
copyrightable content from the moved content. Use BSD-2-Clause license
for new file.
Update #3053.
Update #3875.
Commit 4c09f4b3fe changed the cast from
size_t to ssize_t to try to fix an workspace issue on 16-bit
architectures. This change was reverted by commit
7ff6115b8b for the workspace but not
_Configure_From_workspace(). Change the cast to uintptr_t to be in line
with the Heap Handler.
The CONFIGURE_DISABLE_SMP_CONFIGURATION configuration option and
rtems_configuration_is_smp_enabled() were added during the SMP support
development cycle as a workaround to fix some testsuite failures in SMP
configurations. All use cases were replaced with tests for specific
conditions. The configuration option and test macro were undocumented.
Close#3876.
If the non-preempt mode for threads is supported depends on the
scheduler implementation. Add
_Scheduler_Is_non_preempt_mode_supported() to indicate this.
Update #3876.
Determine the default for CONFIGURE_INIT_TASK_INITIAL_MODES depeding on
whether RTEMS_SMP is defined or not.
In the tests, use CONFIGURE_INIT_TASK_INITIAL_MODES to explicitly
request RTEMS_NO_PREEMPT mode if necessary.
Update #3876.
The IO library used a POSIX key to store an optional POSIX user
environment pointer. This pulled in the POSIX keys support in every
application configuration. Add a user environment pointer to the thread
control block (TCB) instead. Applications which do not need the POSIX
user environment will just get an overhead of one pointer per thread.
Close#3882.
