The alloca() function is machine- and compiler-dependent. For certain
applications, its use can improve efficiency compared to the use of
malloc(3) plus free(3). In certain cases, it can also simplify memory
deallocation in applications that use longjmp(3) or siglongjmp(3).
Otherwise, its use is discouraged.
Signed-off-by: chao an <anchao@lixiang.com>
To support rpmsg ioctl, add cmd_rpmsg function, and update rptun ping to rpmsg ping.
depends on apache/nuttx#11618
Signed-off-by: wangyongrong <wangyongrong@xiaomi.com>
the ping subcommand is guarded with RPTUN_PING in its handler.
its help mesg should have same guard to avoid confusing user.
Signed-off-by: Yanfeng Liu <yfliu2008@qq.com>
This commit corresponds to apache/nuttx#11498 in nuttx.
Move etc romfs mount to sched/init for compatibility with kernel/protected mode.
changes:
- move etc romfs mount from nsh to Nuttx, but keep the script to parse and execute.
- move and rename the related CONFIG, move customized nsh_romfsimg.h to etc_romfs.c
in boards, and no need declaration for romfs_img/romfs_img_len.
Signed-off-by: fangxinyong <fangxinyong@xiaomi.com>
"dns" option of `ifconfig` can work just with `CONFIG_NETDB_DNSCLIENT`,
no need to depend on `CONFIG_NETINIT_DNS` or `CONFIG_NETINIT_DHCPC`.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
Switch the order of setting network mask and gateway, re-order only, no
logic change.
In IPv6 cases, we may set `addr6` and prefix length together to an
interface, but the gateway logic may change the value in `addr6`, and
result in wrong address set to the interface.
The order doesn't change behavior in network stack, so we just need to
make sure the value is correct.
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
Locking the scheduler prior to calling posix_spawn() might lock the
scheduler for a relatively long time, if the file to be loaded is large.
posix_spawn() loads the process into memory, possibly linking it as well
if the binary format is linkable (elf). This can take tens / hundreds of
milliseconds, which basically destroys the OS's real time performance.
Missing the death-of-child signal is a very trivial penalty considering
the alternative.
Simplify the "cmd_codecs_proc" md5 related framework and modify the implementation of the command md5:
Signed-off-by: wangmingrong <wangmingrong@xiaomi.com>
Add the API "nsh_getpid" in "nsh_fsutils. c" and the nsh command pidof. Temporarily support two parameters, - s, and process name
Signed-off-by: wangmingrong <wangmingrong@xiaomi.com>
Add nsh_none to consume all empty traces from nsh.
For the variadic argument case need to add a (inline) function to eat
away the __VA_ARGS__ list, there is no good / portable way to do this by
pre-processor macros, but a function will eat the variadic list whatever
its size is.
Fixes following warning:
CC: grp/lib_getgrbufr.c nsh_timcmds.c: In function 'cmd_date':
nsh_timcmds.c:370:19: warning: variable 'errfmt' set but not used [-Wunused-but-set-variable]
370 | FAR const char *errfmt;
|
Got use-after-free warning under GCC 12 with `-O3` option, and I found
that `nsh_strcat` may realloc `ptr`, then `cmdline` may point to invalid
memory.
Let `cmdline` point to the reallocated `ptr` may solve the problem.
Tested by `alias ll='ls -l'` and `ll /` on sim.
GCC output:
CC: binfmt_unloadmodule.c In function 'nsh_aliasexpand',
inlined from 'nsh_argument' at nsh_parse.c:1879:20:
nsh_parse.c:1196:23: error: pointer 'ptr' used after 'realloc' [-Werror=use-after-free]
1196 | ptr = cmdline + len;
| ~~~~~~~~^~~~~~~~~~~~~~~
In function 'nsh_strcat',
inlined from 'nsh_aliasexpand' at nsh_parse.c:1190:21,
inlined from 'nsh_argument' at nsh_parse.c:1879:20:
nsh_parse.c:1100:27: note: call to 'realloc' here
1100 | argument = (FAR char *)realloc(s1, allocsize);
| ^~~~~~~~~~~~~~~~~~~~~~
Signed-off-by: Zhe Weng <wengzhe@xiaomi.com>
