The commit adds dependency on Kconfig FLASH_PAGE_LAYOUT to subsystems
that really require it:
FCB, NVS, LittleFS
and removes direct selection from '*.conf' files where no longer
needed.
Signed-off-by: Dominik Ermel <dominik.ermel@nordicsemi.no>
Replace the use of a memcmp of a temp buffer filled with
the comparison value with a direct comparison of
the read buffer contents.
Improves speed of the function (no more memset() call)
and reduces stack usage.
Signed-off-by: Mike J. Chen <mjchen@google.com>
NVS lookup cache currently uses CRC8/16 as a hash function
to determine the cache position, which is not ideal choice.
For example, when NVS lookup cache size is 512 and 256
subsequent NVS IDs are written (that is, 0, 1.., 255), this
results in 128 cache collisions.
It is better to use a dedicated integer hash function. This
PR uses one of the 16-bit integer hash functions discovered
with https://github.com/skeeto/hash-prospector project. The
hash function was additionally tested in the context of NVS
lookup cache using simple NVS ID allocation patterns as well
as using real device NVS dump.
Also, add a test case to verify that the hash function is
not extremely bad.
Signed-off-by: Damian Krolik <damian.krolik@nordicsemi.no>
nvs_sector_close has not been initializing nvs_ate.part, before
writing it to flash.
Fixes#58699
Signed-off-by: Dominik Ermel <dominik.ermel@nordicsemi.no>
NVS cache was used only in read function. This commit add the usage of
NVS cache in write and gc functions.
Signed-off-by: Julien D'Ascenzio <julien.dascenzio@paratronic.fr>
During the NVS initialization, if gc had to be done, the NVS cache
rebuild wasn't called.
Signed-off-by: Julien D'Ascenzio <julien.dascenzio@paratronic.fr>
Add CONFIG_CRC for building CRC related routines.
CRC routines are now being built for each application, whether used or
not and are add in the build system unconditionally.
Keep CONFIG_CRC enabled by default for now and until all users have
converted to use the new option.
Partial fix for #50654
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
The NVS data lookup time grows linearly with the number of
allocation table entries to walk through, meaning that if
some data pair in the NVS changes frequently, access to
other data pairs that change rarely can take a lot of time.
It is particularly visible when the NVS is used as the
settings backend since the backend needs to perform multiple
NVS reads to find a requested key.
Implement a simple cache that stores an address of the most
recent ATE for all NVS IDs that fall into the given cache
position. CRC8/16 is used as a hash function used to
distribute NVS IDs across the cache entries.
The cache entries are only invalidated when an NVS sector
is erased as part of the garbage collector task.
Signed-off-by: Damian Krolik <damian.krolik@nordicsemi.no>
In order to bring consistency in-tree, migrate all subsystems code to
the new prefix <zephyr/...>. Note that the conversion has been scripted,
refer to zephyrproject-rtos#45388 for more details.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Add a new API call to replace nvs_init: nvs_mount. The new API does the
same as nvs_init except that it assumes to be provided with a valid
flash device via `struct nvs_fs` `flash_device` field. Previously, it
was not possible to avoid the runtime overhead of device_get_binding()
even if the flash device was known at compile time.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
A badly written sector close ate that has correct crc8 could allow
jumps outside the assigned flash area. This behavior is fixed.
A possibility existed that a badly erased sector or a incomplete write
of a large item created a empty closed sector. This has been fixed by:
a. Erase verification.
b. Clearing such a sector at startup.
Fixes#34382
Signed-off-by: Laczen JMS <laczenjms@gmail.com>
Co-Authered-by: Andrzej Puzdrowski <andrzej.puzdrowski@nordicsemi.no>
Document the fact than NVS image of different wbs
are compatible each other if psychical ATE size
is kept.
Signed-off-by: Andrzej Puzdrowski <andrzej.puzdrowski@nordicsemi.no>
Signed-off-by: Damian Krolik <damian.krolik@nordicsemi.no>
When existing records stored in the NVS are not properly
aligned according to the current flash driver requirements,
fs->data_wra may be initialized with an unaligned address.
Fix the initialization code, so that fs->data_wra is rounded
up to the nearest multiple of the current flash driver block
size.
The situation may occur during a firmware upgrade which
introduces a new flash driver or changes its parameters.
Signed-off-by: Damian Krolik <damian.krolik@nordicsemi.no>
Removed two unneeded `if` statements. These statements had not
effect in function results.
Signed-off-by: Andrzej Puzdrowski <andrzej.puzdrowski@nordicsemi.no>
flash_write_protection_set() was deprecated.
This patch removes usage of it and the test case which was
testing deprecated API behaviors.
Signed-off-by: Andrzej Puzdrowski <andrzej.puzdrowski@nordicsemi.no>
The second argument of this foonction is a bool, so passing 0 and 1
is incorrect.
Coccinelle script:
@@
expression e;
@@
flash_write_protection_set(e,
(
- 0
+ false
|
- 1
+ true
)
)
Signed-off-by: Ievgenii Meshcheriakov <ievgenii.meshcheriakov@nordicsemi.no>
nvs_gc does not verify the crc8 of close_ate before using
close_ate.offset. This means that close_ate.offset could contain an
offset that points beyond valid ate's in the sector. For example, there
might be a valid ate at offset 0x100 but close_ate.offset is 0x200.
If that is the case that value will not be moved and so it will be lost.
Solve this by refactoring the recovery loop from nvs_prev_ate into
nvs_recover_last_ate and use that function in nvs_gc if a corrupt
close_ate is found.
The crc8 of gc_ate is not checked before trying to find another ate
with the same id. If there are no valid ate with that id in the whole
fs the inner while(1)-loop will never stop since the break condition
includes a check for a correct crc8.
Solve this by skipping gc_ate's with an invalid crc8.
Fixes#26407
Signed-off-by: Tobias Svehagen <tobias.svehagen@gmail.com>
Pointer to flash_parameters have been added to nvs_fs structure and it
is no longer needed to store write_block_size within the nvs_fs.
Signed-off-by: Dominik Ermel <dominik.ermel@nordicsemi.no>
The current code truncates the size_t block size to a u8_t, which
causes a 256 or 512 byte block size to be truncated to zero and
passing the size check.
Check as a size_t then truncate once the check passes.
Signed-off-by: Michael Hope <mlhx@google.com>
Replace all occurences of BUILD_ASSERT_MSG() with BUILD_ASSERT()
as a result of merging BUILD_ASSERT() and BUILD_ASSERT_MSG().
Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
This reverts commit 8739517107.
Pull Request #23437 was merged by mistake with an invalid manifest.
Signed-off-by: Carles Cufi <carles.cufi@nordicsemi.no>
Replace all occurences of BUILD_ASSERT_MSG() with BUILD_ASSERT()
as a result of merging BUILD_ASSERT() and BUILD_ASSERT_MSG().
Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
In nvs writing addresses are u32_t. Coverty reports two situations
where the address could be converted (unwanted) to a signed value.
Both have been corrected.
There is however a general problem with flash API where the addresses
are defined as off_t which is a s32_t. These are converted in the flash
hal to u32_t. As a result of this only half of the possible range can
be used.
Solves #20867 and #20866
Signed-off-by: Laczen JMS <laczenjms@gmail.com>
Use this short header style in all Kconfig files:
# <description>
# <copyright>
# <license>
...
Also change all <description>s from
# Kconfig[.extension] - Foo-related options
to just
# Foo-related options
It's clear enough that it's about Kconfig.
The <description> cleanup was done with this command, along with some
manual cleanup (big letter at the start, etc.)
git ls-files '*Kconfig*' | \
xargs sed -i -E '1 s/#\s*Kconfig[\w.-]*\s*-\s*/# /'
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
Unused since commit 7d2e59813f ("subsys: fs/nvs: Rewrite for improved
robustness").
Found with a script.
Signed-off-by: Ulf Magnusson <Ulf.Magnusson@nordicsemi.no>
When overwriting an NVS item with data that was a truncated version of
the existing data, the "is this already saved" logic was ignoring the
differing lengths and not saving the new item because the data matched.
Fixes#19250
Signed-off-by: Justin Brzozoski <justin.brzozoski@signal-fire.com>
Make sure that the last entry added is deleted correctly by storing the
fact that one was found in a local variable.
Fix by Laczen JMS <laczenjms@gmail.com>
Fixes#18813.
Signed-off-by: Carles Cufi <carles.cufi@nordicsemi.no>
No need to write deletion entry for non-existing entry.
Furthermore such a write to filled up storage will make another
writes impossible as it occupied delete nvs_ate space, but it
does not remove anything.
Signed-off-by: Andrzej Puzdrowski <andrzej.puzdrowski@nordicsemi.no>
During initialization NVS analyzes open sector for
find write addresses. It was possible to infinite loop
in case when there was no space for any new ATE in the sector.
This patch introduce check for space available in the sector which
fix the issue.
fixes#17891
Signed-off-by: Andrzej Puzdrowski <andrzej.puzdrowski@nordicsemi.no>
NVS always leave space for deletion ate in a full sector.
Even that it was not possible to write delete nvs_ate to the
full sector. Because of that it was possible to fill up NVS,
and be able to delete nothing.
This patch introduce recognition of case
the delete ate is writing, and allow to write such nvs_ate to
opened full sector.
Signed-off-by: Andrzej Puzdrowski <andrzej.puzdrowski@nordicsemi.no>
Previous ATE searching is accelerate by reading sector
close ate, which allow to skip reading whole storage.
ATE is already covered by crc8. That patch introduce
check on correctness offset read form close ate. This
increases meta-data integrity check level.
This also preserves against possible looping when read
incorrect or not a NVS storage image.
fixes#16899
Signed-off-by: Andrzej Puzdrowski <andrzej.puzdrowski@nordicsemi.no>
The sector size is 0 will pass "fs->sector_size % info.size" then start
a loop in nvs_startup() and never return. So retrun an error if the
sector size is 0.
Signed-off-by: Harry Jiang <explora26@gmail.com>
move flash.h to drivers/flash.h and
create a shim for backward-compatibility.
No functional changes to the headers.
A warning in the shim can be controlled with CONFIG_COMPAT_INCLUDES.
Related to #16539
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
move crc.h to sys/crc.h and
create a shim for backward-compatibility.
No functional changes to the headers.
A warning in the shim can be controlled with CONFIG_COMPAT_INCLUDES.
Related to #16539
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
move nvs/nvs.h to fs/nvs.h and
create a shim for backward-compatibility.
No functional changes to the headers.
A warning in the shim can be controlled with CONFIG_COMPAT_INCLUDES.
Related to #16539
Signed-off-by: Anas Nashif <anas.nashif@intel.com>
