after below change merge to kernel, spin_lock() will turn off preemption by default,
but this change is not applicable to all scenarios. The locations in the kernel that
use spin_lock() extensively only require short critical sections and do not trigger
scheduling, which leads to serious performance degradation of NuttX in AMP mode.
In this PR, I try to expose similar problems and hope that each subsystem will carefully check the code coverage
https://github.com/apache/nuttx/pull/14578
|commit b69111d16a
|Author: hujun5 <hujun5@xiaomi.com>
|Date: Thu Jan 23 16:14:18 2025 +0800
|
| spinlock: add sched_lock to spin_lock_irqsave
|
| reason:
| We aim to replace big locks with smaller ones. So we will use spin_lock_irqsave extensively to
| replace enter_critical_section in the subsequent process. We imitate the implementation of Linux
| by adding sched_lock to spin_lock_irqsave in order to address scenarios where sem_post occurs
| within spin_lock_irqsave, which can lead to spinlock failures and deadlocks.
|
| Signed-off-by: hujun5 <hujun5@xiaomi.com>
Signed-off-by: chao an <anchao.archer@bytedance.com>
reason:
We decouple semcount from business logic
by using an independent counting variable,
which allows us to remove critical sections in many cases.
Signed-off-by: hujun5 <hujun5@xiaomi.com>
* Make readv/writev implementations update struct uio
This can simplify partial result handling.
* change the error number on the overflow from EOVERFLOW to EINVAL
to match NetBSD
* add a commented out uio_offset field. I used "#if 0" here as
C comments can't nest.
* add a few helper functions
Note on uio_copyfrom/uio_copyto:
although i'm not quite happy with the "offset" functionality,
it's necessary to simplify the adaptation of some drivers like
drivers/serial/serial.c, which (ab)uses the user-supplied buffer
as a line-buffer.
Summary:
There are differences in different server implementations, so it is necessary to check whether the returned nwqid satisfies the requested number.
Signed-off-by: chenrun1 <chenrun1@xiaomi.com>
Fixes an issue in kernel build where the user addresses passed
to accept() would be accessed when the wrong MMU mappings were
active. A crash would manifest when attempting to accept() on a
TCP server socket for instance under significant load. The accept
event handler would be called by the HP worker upon client
connection. At this point, accept_tcpsender() would attempt to
write to `addr` resulting in a page fault. Reproducibility would
depend on the current system load (num tasks or CPU stress) but
in loaded environments, it would crash almost 100% of the times.
It should be noted that Linux does this the other way around: it
operates on kernel stack allocated data and once done, it copies
them to user. This can also be a viable alternative, albeit with
one extra copy and a little extra memory.
Signed-off-by: George Poulios <gpoulios@census-labs.com>
reason:
The list->fl_files may be modified by files_extend.
If it is being modified, there will be a race condition for fl_files[l1][l2].
Signed-off-by: hujun5 <hujun5@xiaomi.com>
Check fs_fatsecperclus' value when read from eMMC device. Return an error if it is zero.
Signed-off-by: Yinzhe Wu <Yinzhe.Wu@sony.com>
Reviewed-by: Jacky Cao <Jacky.Cao@sony.com>
Tested-by: Yinzhe Wu <Yinzhe.Wu@sony.com>
POSIX requires that the shm objects are zero-initialized. This has been broken
in some earlier commits (starting from 9af5fc5d09)
Also fix the flat build memory allocation to allocate both object data and payload
in the same chunk (as the comment also suggests). This saves allocations and memory
in a system with lots of shm objects.
Signed-off-by: Jukka Laitinen <jukkax@ssrc.tii.ae>
Most of developers will not have the scenarios of open/close file descriptors in multi-threads,
The default option will incur additional code size overhead for such devices.
this PR will preserve the behavior before PR#13296 was introduced, and ensure that the default code size is not affected.
Note that this option will ensure the safety of access to the file
system from multi-tasks (Task A blocking rw(fd), then Task B close(fd)),
the disadvantage is that it will increase the amount of code-size,
there is no need to enable this option if the application could ensure
he file operations are safe.
Signed-off-by: chao an <anchao@lixiang.com>
After tcb is destroyed, it is very dangerous to back reference tcb through file.
This commit will perform file operations while ensuring the validity of tcb during
fsync, with will avoid tcb check in each subsystem.
Signed-off-by: chao an <anchao@lixiang.com>
Modify the kernel to use only atomic_xx and atomic64_xx interfaces,
avoiding the use of sizeof or typeof to determine the type of
atomic operations, thereby simplifying the kernel's atomic
interface operations.
Signed-off-by: zhangyuan29 <zhangyuan29@xiaomi.com>
Because they are not expected to apply cleanly.
In that case, disable all features which require the local patches.
Note: lfs_util.patch is actually necessary regardless of
the littlefs version. It should be ok to drop it for the simplest
flat memory model configurations with a single heap though. I plan
to fix it once littlefs 2.9.4 is released.
Tested with esp32s3-devkitc:toywasm + CONFIG_FS_LITTLEFS_VERSION="v2.9.3".
Panic occured if no txtable was found in last eraseblock of MTD device and `CONFIG_TXTABLE_DEFAULT_PARTITION` was disabled
Backtrace
sched_dumpstack
/workspace/nuttx/libs/libc/sched/sched_dumpstack.c:71
dump_assert_info
/workspace/nuttx/sched/misc/assert.c:718
xtensa_user_panic
/workspace/nuttx/arch/xtensa/src/common/xtensa_assert.c:188 (discriminator 1)
xtensa_user
??:?
_xtensa_user_handler
/workspace/nuttx/arch/xtensa/src/common/xtensa_user_handler.S:194
??
??:0
parse_txtable_partition
/workspace/nuttx/fs/partition/fs_txtable.c:136
parse_partition
/workspace/nuttx/fs/partition/fs_partition.c:165
parse_mtd_partition
/workspace/nuttx/fs/partition/fs_partition.c:291
Signed-off-by: wangjianyu3 <wangjianyu3@xiaomi.com>
Hopefully, It's more user-friendly than the current method,
which is just an overridable make variable. It would also help
to make some features (and their associated local patches)
conditional on littlefs version in future.
Also, update CMakeLists.txt to use the same littlefs version as Makefile.
reason:
When entering an exception or interrupt, there are two sets of registers:
one is the "running regs", which we need to save,
and the other is the "ready to running regs", which we may soon use.
For consistency in logic, we can always store the "running regs" in the regs field of g_running_tasks,
otherwise it may lead to errors in the storage location of the "running regs."
When we need to access the "running regs," we should uniformly retrieve them from the regs field of g_running_tasks.
As the next step, we will rename the set_current_regs/up_current_regs functions
for each architecture to more appropriate names, solely for the purpose of identifying interrupts.
Signed-off-by: hujun5 <hujun5@xiaomi.com>
====================================================================================
Configuration/Tool: spresense/example_camera,CONFIG_ARM_TOOLCHAIN_GNU_EABI
2024-11-12 12:24:37
------------------------------------------------------------------------------------
Cleaning...
Configuring...
Disabling CONFIG_ARM_TOOLCHAIN_GNU_EABI
Enabling CONFIG_ARM_TOOLCHAIN_GNU_EABI
Building NuttX...
mount/fs_automount.c: In function 'automount_initialize':
Error: mount/fs_automount.c:816:7: error: 'return' with no value, in function returning non-void [-Werror=return-type]
816 | return;
| ^~~~~~
In file included from mount/fs_automount.c:43:
/github/workspace/sources/nuttx/include/nuttx/fs/automount.h:176:11: note: declared here
176 | FAR void *automount_initialize(FAR const struct automount_lower_s *lower);
| ^~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
Signed-off-by: chenrun1 <chenrun1@xiaomi.com>
