Running check may be slow and increase result time in the other of 70ms
for the list method.
Add an option to skip running check by passing the skip_running_check as
true in the list method args to speedup execution if the running info is
not needed.
With the option set to true the execution time lower to just 5ms.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Running check is supported only in procd scripts. This cause prolonged
execution time since the function needs to timeout.
To fix this check if the script USE_PROCD and run running check only if
supported.
Also provide running info only if the running check is supported.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Currently we stop searching at the first occurence of START or STOP
entry. This is wrong since we totally miss the other data (START or
STOP) in the occurence of the other.
Fix and improve script scanning by:
- Increase the line max length to 255 char to read it in one go.
- Scan only the first 10 lines.
- Don't stop at the first occurence and try to search also for the other
data.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
So that consumers don't have to fiddle around with mapping frequencies
to bands, which everyone seems to do a little differently.
Signed-off-by: Andre Heider <a.heider@gmail.com>
A channel can declare restriction where it should be used only indoors
or should not be used at all.
Expose these restriction in the channel data as additional info with the
restricted info.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Don't hardcode bit/name pairs, instead iterate over what's known to the
library and use that instead.
This automatically adds the missing ciphers CCMP256 and GCMP256 - and any
future ones.
The only difference in the output is the order of the 'hwmodes' array.
Signed-off-by: Andre Heider <a.heider@gmail.com>
Write unhandled exceptions in invoked ucode method handlers to stderr in
order to simplify debugging failing scripts.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Ensure that the default module search path is initialized before compiling
ucode handler scripts in order to support compile time import statements.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
gcc 10 with -O2 reports following:
In function ‘strncpy’,
inlined from ‘rpc_sys_packagelist’ at /opt/devel/openwrt/c-projects/rpcd/sys.c:244:4:
/usr/include/x86_64-linux-gnu/bits/string_fortified.h:106:10: error: ‘__builtin_strncpy’ specified bound 128 equals destination size [-Werror=stringop-truncation]
106 | return __builtin___strncpy_chk (__dest, __src, __len, __bos (__dest));
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In function ‘strncpy’,
inlined from ‘rpc_sys_packagelist’ at /opt/devel/openwrt/c-projects/rpcd/sys.c:227:4:
/usr/include/x86_64-linux-gnu/bits/string_fortified.h:106:10: error: ‘__builtin_strncpy’ specified bound 128 equals destination size [-Werror=stringop-truncation]
106 | return __builtin___strncpy_chk (__dest, __src, __len, __bos (__dest));
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Since it is not possible to avoid truncation by strncpy, it is necessary
to make sure the result of strncpy is properly NUL-terminated and the
NUL must be inserted explicitly, after strncpy has returned.
References: #10442
Reported-by: Alexey Smirnov <s.alexey@gmail.com>
Signed-off-by: Petr Štetiar <ynezz@true.cz>
It makes little sense to execute rpcd ucode plugin scripts in template
mode since those scripts are supposed to output structured JSON data,
so change the parse config to compile scripts in raw mode.
Also initialize the default library search path which is required in
recent ucode versions.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Fix compilation error with GCC 12.
In file included from /home/ansuel/openwrt/staging_dir/target-aarch64_cortex-a53_musl/usr/include/libubus.h:23,
from iwinfo.c:21:
In function 'blobmsg_close_array',
inlined from 'rpc_iwinfo_assoclist' at iwinfo.c:643:3:
/home/ansuel/openwrt/staging_dir/target-aarch64_cortex-a53_musl/usr/include/libubox/blobmsg.h:250:9: error: 'c' may be used uninitialized [-Werror=maybe-uninitialized]
250 | blob_nest_end(buf, cookie);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~
iwinfo.c: In function 'rpc_iwinfo_assoclist':
iwinfo.c:564:15: note: 'c' was declared here
564 | void *c, *d, *e;
| ^
cc1: all warnings being treated as errors
ninja: build stopped: subcommand failed.
Signed-off-by: Christian 'Ansuel' Marangi <ansuelsmth@gmail.com>
The public libucode api has been revised to return and expect an
uc_program_t pointer instead of a main function reference.
The program (former main function) is also not implicitly released
by uc_vm_execute() anymore.
Adjust the ucode plugin accordingly to match the new requirements.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
The rpcd ucode plugin allows utilizing ucode scripts to register ubus
objects and to implement the objects method callbacks.
Upon startup, rpcd will compile and execute each ucode script in
`$INSTALL_PREFIX/share/ucode/` and register ubus proxy objects and
methods definitions according to the signature returned by the script.
Refer to examples/ucode/example-plugin.uc for details of the signature
format.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Drop the historically inherited "luci-rpc-" prefix and replace it with a
more appropriate "rpcd-plugin-" string.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Search plugins, ACL ressources and other data files relative to the compile
time CMAKE_INSTALL_PREFIX instead of hardcoding `/usr`.
Also ensure that plugin libraries are installed into the lib/rpcd subdir.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
We need to use blobmsg_get_u64 for RPC_DUMP_EXPIRES as this is now
BLOBMSG_TYPE_INT64.
Fixes: 20bf958bea ("session: use uloop_timeout_remaining64")
Signed-off-by: Stijn Tintel <stijn@linux-ipv6.be>
Acked-by: Jo-Philipp Wich <jo@mein.io>
Currently, rpcd cached /etc/config/rpcd once on the first login and then
keeps checking login attempts and ACL settings against the in-memory
copy until the process is restarted.
This is not desirable as we would like added, modified or removed accounts
to be effective immediately. In order to achive this behaviour, ensure to
unload the currently loaded rpcd configuration before looking up the login
account section.
Ref: https://github.com/openwrt/luci/issues/5179
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Expose 802.11ax HE rate as well as HW / HT mode information. This is
required to add 802.11ax support to LuCI.
Signed-off-by: David Bauer <mail@david-bauer.net>
Having the ABI version appended to the package name is not very helpful
when checking for updated versions of a package online or requesting
a new image for the device from an updater service. In both cases,
the ABI version could have been bumped meanwhile and that name is then
simply not found.
Resolve this by removing the appended ABI version in the output of the
rpc-sys packagelist call.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
This is required to avoid potential use-after-free errors through the
uci_set()->uci_delete()->uci_expand_ptr() call chain when passing
zero-length strings as values.
Ref: https://bugs.openwrt.org/index.php?do=details&task_id=3528
Suggested-by: olegio170 <olegios170@gmail.com>
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Iwinfo export the center channel sued by the wifi. Include this data in
the ubus info call to better know the channel utilizzation of the wifi.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
Iwinfo exports ht and vht operation info useful to get channel info of
nearby stations. Add these new info to ubus output.
Signed-off-by: Ansuel Smith <ansuelsmth@gmail.com>
The special characters '!' and '*' in /etc/shadow are
used to disable password login for a specific account.
The character 'x' has no special meaning, but should not
be interpreted as an empty password.
However, rpcd did treat these special characters like no
password was set, which allows access even though the account
is disabled.
By removing the additional checks for these characters, the
encrypted password string is passed to crypt, which returns NULL
if the salt has an invalid format and therefore access is denied.
Fixes: FS#2634
Signed-off-by: Fabian Bläse <fabian@blaese.de>
When the initial exec permission check on the executable path fails,
concatenate the command line with spaces and use the resulting string
as lookup path for a second exec permission check.
This allows for exec acls similar to this example:
"file": {
"/usr/bin/program --flag --option=1 arg *": [ "exec" ]
}
The example above would allow executing `/usr/bin/program` with the
arguments `--flag`, `--option=1` and `arg` in exactly this order,
followed by any number of optional arguments as denoted by the
asterisk.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
When parsing the JSON output of exec plugins, store integer values exceeding
32bit value limits as 64bit integer blob values.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Prevent a theoretical leak of the args memory when the executable path
cannot be found.
Signed-off-by: Yousong Zhou <yszhou4tech@gmail.com>
[fix whitespace, commit description]
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
A previous commit changed the allocation method for the call context to
include the method name string memory directly, so we must not treat the
method member separately anymore.
Fixes: 37aa919 ("plugin: fix leaking invoked method name for exec plugins")
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
In some cases, e.g. when subsequently setting multiple empty option
values, uci_set() might free the section pointer of the given reused
uci_ptr structure without zeroing it, leading to a use-after-free on
processing subsequent options.
Avoid this issue by clearing the lookup pointer flags in order to
prevent uci_set() from incorrectly branching into a uci_delete()
operation leading to the freeing of the section member.
Ref: http://lists.infradead.org/pipermail/openwrt-devel/2019-October/019592.html
Reported-by: Daniel Danzberger <daniel@dd-wrt.com>
Suggested-by: Yousong Zhou <yszhou4tech@gmail.com>
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
The invoked method name was separately duplicated from the call_context
structure. The structure itself is eventually freed by rpc_exec_reply()
but the method string it points to is lost after that.
Use calloc_a() instead to allocate the string copy buffer together with
the context structure, to ensure that all involved memory is freed.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
Some sysfs or proc files contain more than 128 byte of data, e.g. the
/proc/filesystems or /proc/mounts files.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
