| Age | Commit message (Collapse) | Author |
|---|
|
Both input and output callbacks for uloop tasks are optional, but the
low level io callback implementation did not properly deal with an absent
ucode output callback, triggering an exception in managed code due to
invoking a null value as function.
Fix this issue by checking for the availability of a callable output
function and simply discarding the received task message otherwise.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Prevent GC'ing (and thus tearing down) an active ubus connection as long as
there's still unfinished deferred request contexts alive.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Some ubus users require access to the original ubus error status returned
by various operations for fine grained error handling.
Extend the error() function with an optional boolean argument which causes
the function to return the numeric error code instead of a preformatted
message when invoked.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Add constants for ubus status codes to the toplevel module scope in order
to avoid the need for magic values in the code.
Suggested-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
The implicit return style for sending ubus method replies currently always
emits an UBUS_STATUS_NO_DATA code in case neither req.reply() was called,
nor a deferred or object were returned by the handler function.
This slightly complicates the implementation of handlers that do not wish
to send reply data but simply acknowledge the request with an
UBUS_STATUS_OK code.
In order to simplify this use case, allow handlers to override the default
status by treating integer return values as ubus error codes.
After this change, the following handler:
function (request) {
/* do some work */
request.reply(null, 0);
}
... can be rewritten as:
function (request) {
/* do some work */
return 0;
}
Suggested-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Rework the `fs.open()` implementation to accept an optional third file
permission argument which is applied to newly created files.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
When building with gcc-10 and -O2, the following warning in ubus.c is
triggered during the compilation:
In function ‘uc_ubus_object_register’,
inlined from ‘uc_ubus_publish’ at .../ubus.c:1521:10:
.../ubus.c:1464:14: error: ‘strncpy’ output truncated before terminating nul copying as many bytes from a string as its length [-Werror=stringop-truncation]
1464 | obj->name = strncpy(onptr, ubus_object_name, namelen);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.../ubus.c: In function ‘uc_ubus_publish’:
.../ubus.c:1447:12: note: length computed here
1447 | namelen = strlen(ubus_object_name);
| ^~~~~~~~~~~~~~~~~~~~~~~~
cc1: all warnings being treated as errors
Solve this issue by using memcpy() instead. We already take care of
allocating a zeroed, strlen() + 1 sized destination buffer so loosing
the `\0` byte of the source string is perfectly fine.
Fixes: #100
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Some pecularities in the encoding of the IFLA_AF_SPEC attribute make it
unsuitable for table driven parsing/generation.
To solve this issue, introduce specific datatype handling for IFLA_AF_SPEC
and parse/generate the RTA depending on the address family of the containing
netlink message.
Also add some missing constants while we're at it.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
ubus: end uloop on exceptions in managed code
|
|
Instead of silently continuing, end the uloop when encountering exceptions
in ucode callbacks to let those exceptions propagate to the host program.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Instead of silently continuing, end the uloop when encountering exceptions
in ucode callbacks to let those exceptions propagate to the host program.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Decode IFLA_STATS64 attribute and make contained counters available to ucode.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
For certain operations, such as bringing up interfaces, it is required to
initialize the ifi_change mask in the ifinfomsg struct.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
rtnl: update NETLINK_GET_STRICT_CHK socket flag with every request
|
|
So far the NETLINK_GET_STRICT_CHK socket flag was only set on the implicit
socket creation performed during the first request and ignored for
subsequent ones which made it impossible to perform only some requests
with enabled strict checking.
Modify the logic to check the flag state for every request and change
it if needed. This allows performing both strict and non-strict requests
over the same connection.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Report the noise value as signed integer to calling ucode.
Reported-by: John Crispin <john@phrozen.org>
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Extend the `unpack()` function to take an optional, second offset parameter
which is useful to skip an initial portion of the input data without having
to encode pad bytes into the format string.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
rtnl: fix segmentation fault on parsing linkinfo RTA without data
|
|
rtnl: zero request message headers
|
|
rtnl: fix premature netlink reply receive abort
|
|
Some link types, such as veth, yield an IFLA_LINKINFO nla without an
embedded IFLA_INFO_DATA / INFLA_INFO_SLAVE_DATA nla which causes the
nla converter to dereference a NULL nla pointer.
Properly deal with such cases and check for the existence of the
child nla before attempting to parse it.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
rtnl: avoid stray "netlink: %d bytes leftover after parsing attributes."
|
|
For route netlink request messages having a header struct, uc_nl_request()
invokes nlmsg_reserve() to reserve room for the struct data but the
nlmsg_reserve() function only zeroes additional alignment bytes, not the
actual reserved buffer space.
Extend the existing logic to explicitly zero out the reserved header space
in order to avoid sending uninitialized struct member values to the kernel.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
The nl_recvmsgs() logic in uc_nl_request() incorrectly stopped reading
the socket before the netlink ACK message was handled for non-multipart
replies.
This caused subsequent requests to incorrectly receive the ACK of the
previous request, leading to a failure to receive the actual reply.
Fix this issue by continue reading the socket until either the finish
callback for multipart (dump) messages or the ack callback for non-
multipart messages was received.
This fix is basically the same as the one applied to the nl80211 module
in 54ef6c0 ("nl80211: fix premature netlink reply receive abort").
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Some nested RTAs such as IFLA_INET_CONF do not contain actual sub-RTAs but
just an array of integers. Avoid calling a no-op `nla_parse()` for such
attributes to suppress the non-harmful leftover bytes warning emitted by
libnl.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
When packing a format such as `!6C*`, the `*` format was not included into
the result buffer due to improper tracking of the function argument offset.
Solve this issue by taking field repetitions into account when tracking
the argument offset while calculating the size of dynamic arguments.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
If the read limit passed to fs.readfile() is smaller than BUFSIZ then
disable stdio input buffering to avoid overreading the underlying file.
This is useful when reading small amounts of data from special files
such as /dev/urandom, where a readfile call limited to 16 bytes might
actually read 4096 due to stdio buffering.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
The `fs.readfile()` and `fs.writefile()` functions allow efficient reading
and writing of whole files, reducing the required boilerplace code compared
to using the `open()`/`read()`/`write()`/`close()` API.
- `fs.readfile()` takes two arguments; the path to open and an optional
limit value. If limit is omitted, `null` or negative, the entire file
contents are read, otherwise the specified amount of bytes. Returns the
read file contents as string or `null` on error.
- `fs.writefile()` takes three arguments; the path to open/create, the
contents to write and an optional limit value. If limit is omitted, the
entire content is written, otherwise just the specified amount of bytes.
Non-string content arguments are internally converted to strings, `null`
is treated as empty string. Returns the amount of bytes written or `null`
on error.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Make sure fs.dirname() doesn't truncate the last character of the
returned path. Previously ucv_string_new_length was called with a
length which no longer included the last character (which had just
been tested not to be a '/' or '.' and hence broke the loop at that
point).
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
[testcase added]
Signed-off-by: Paul Spooren <mail@aparcar.org>
[testcase folded into this commit and fixed]
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Tasks are similar to processes but instead of executing a new process, an
ucode function is invoked instead, running independently of the main
process.
Example usage:
uloop.init();
let t = uloop.task(
// program function
function(pipe) {
let input = pipe.receive();
pipe.send({ got_input: input });
return { result: true };
},
// parent recv function, invoked when task function calls pipe.send()
function(res) {
printf("Received output message: %.J\n", res);
},
// parent send function, invoked when task function calls pipe.receive()
function() {
let input = { test: "Example" };
printf("Sending input message: %.J\n", input);
return input;
}
);
uloop.run();
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
OS X `socket()` does not support the `SOCK_NONBLOCK` or `SOCK_CLOEXEC`
constants, so apply these flags using `fcntl()` after creating the socket.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
It should be enough to include `sys/types.h` to obtain `major()` and
`minor()` definitions on OS X, so avoid including the the Linux specific
`sys/sysmacros.h` header in this case.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Since `execvpe()` is a GNU extension, fall back to using `execve()` on OS X.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Suggested-by: John Crispin <john@phrozen.org>
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Extend the ubus binding to cover ubus event handling APIs.
Instantiating ubus event listener:
listener = conn.listener(
"event.type.*",
function (type, data) {
...event callback...
}
);
listener.remove();
Broadcasting events:
conn.event("event.type.foo", { ...event data... });
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
- Add more detailled error messages
- Introduce helpers for fetching and validating function call arguments
- Get rid of some uneeded blob->jso->ucv conversions
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Extend the ubus binding to cover ubus object publishing, notifications on
objects, as well as subscriber APIs.
Instantiating ubus objects:
obj = conn.publish("objname", {
methodname: {
args: { ...argspec... },
call: function(request) { ...method handler... }
},
...
}, function() { ...subscription status change handler... });
obj.notify(...);
obj.remove();
Emitting notifications:
obj.notify("notificationtype", { ...notification data... },
function(type, data) { ...data callback... },
function(idx, ret) { ...status callback... },
function() { ...completion callback... },
100 /* timeout */
);
Instantiating subscribers:
sub = conn.subscriber(
function(notify) { ...notification handler... },
function(id) { ...object gone handler... }
);
sub.subscribe("objname");
sub.unsubscribe("objname");
sub.remove();
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
In some cases, errno contains stale values from prior function invocations
which might lead to random failures in uc_uloop_run(), uc_uloop_timer_set()
and uc_uloop_timer().
Solve this issue by explicitly initializing errno to 0.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
The uloop module allows controlling the uloop event loop and supports adding
timeouts, processes and file descriptors.
Example:
#!ucode -RS
let fs = require("fs");
let uloop = require("uloop");
let fd1 = fs.popen("echo 1; sleep 1; echo 2; sleep 1; echo 3", "r");
let fd2 = fs.popen("echo 4; sleep 1; echo 5; sleep 1; echo 6", "r");
function fd_read_callback(flags) {
if (flags & uloop.ULOOP_READ) {
let line = this.handle().read("line");
printf("Line from fd <%s/%d>: <%s>\n",
this, this.fileno(),
trim(line));
}
}
uloop.init();
uloop.timer(1500, function() {
printf("Timeout after 1500ms\n");
});
uloop.handle(fd1, fd_read_callback, uloop.ULOOP_READ);
uloop.handle(fd2, fd_read_callback, uloop.ULOOP_READ);
uloop.process("date", [ "+%s" ], { LC_ALL: "C" }, function(exitcode) {
printf("Date command exited with code %d\n", exitcode);
});
uloop.run();
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
The `access()` function allows testing the path given in the first argument
for accessibility according to the permissions specified in the second
mode string argument. The mode string characters may be `r`, `w`, `x` or
`f` which correspond to `R_OK` - path is readable, `W_OK` - path is
writable, `X_OK` - path is executable or `F_OK` - path exists respectively.
The `mkstemp()` function creates a secure temporary file, unlinks it and
returns the open file handle. The temporary path is constructed based on
the optional template argument. If the template argument contains a slash,
the path is taken as-is, if it contains no slashes, `/tmp/` is prepended.
If the template does not end with `XXXXXX`, a `.XXXXXX` suffix is appended
to the path. If the template is omitted, `/tmp/XXXXXX` is used.
The `file.flush()` and `proc.flush()` functions call `fflush()` on the
underlying file handle respectively. They take no arguments.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Implement a new `*` format which acts like `s` on unpack but accepts
input records which are shorter than the specified length, e.g. the
following call will yield "abc" while an equivalent "10s" format would
fail:
unpack("2*", "abc") // [ "ab" ]
unpack("10*", "abc") // [ "abc" ]
unpack("10s", "abc") // null
The `*` format is primarily useful to extract the remainder of a variable
length record without having to encode the specific length of the record
directly into the format string.
When packing records, the `*` format takes at most as many bytes as
specified in the format string repeat count. If the input string is
shorter than the given repeat count, only as many bytes as present in
the input string are taken. A bare `*` without any repeat count will take
all bytes from the given input string:
pack("2*", "abc") // "ab"
pack("10*", "abc") // "abc"
pack("*", "abc") // "abc"
pack("10s", "abc") // "abc\0\0\0\0\0\0\0"
Additionally prevent invalid memory accesses when unpacking a buffer
shorter than the length expected by the format string.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
This avoid triggering -Werror=unused-result compilation failures with
certain toolchains.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Use uc_pack_double() and uc_unpack_double() from core to avoid unnecessary
code duplication.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
- Parse integer literals as unsigned numeric values in order to be able
to represent the entire unsigned 64bit value range
- Stop parsing minus-prefixed integer literals as negative numbers but
treat them as separate minus operator followed by a positive integer
instead
- Only store unsigned numeric constants in bytecode
- Rework numeric comparison logic to be able to handle full 64bit
unsigned integers
- If possible, yield unsigned 64 bit results for additions
- Simplify numeric value conversion API
- Compile code with -fwrapv for defined signed overflow semantics
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Implement support for opening existing file descriptors as well as acquiring
the descriptor number from open process and file handles.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Introduce a new ubus.defer() method which initiates asynchroneous requests
and invokes the completion callback passed as 4th argument once the reply
is received.
This allows multiplexing mutliple ubus requests with the same ucode VM
context / the same uloop event loop.
In case the ucode context is not running under an active uloop, the ubus
module will spawn uloop itself once the first asynchroneous request is
launched and terminate the loop once all pending requests finished.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
The nl_recvmsgs() logic in uc_nl_request() incorrectly stopped reading
the socket before the netlink ACK message was handled for non-multipart
replies.
This caused subsequent requests to incorrectly receive the ACK of the
previous request, leading to a failure to receive the actual reply.
Fix this issue by continue reading the socket until either the finish
callback for multipart (dump) messages or the ack callback for non-
multipart messages was received.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Add three new functions `dirname()`, `basename()` and `lsdir()`.
The `basename()` and `dirname()` functions behave like their libc
counterparts and return the filename and directory portion of a
given path respectively. If the path argument is missing or not
a string, null is returned.
Examples:
dirname("/usr/lib") -> "/usr"
dirname("/usr/") -> "/"
dirname("usr") -> "."
dirname("/") -> "/"
dirname(".") -> "."
dirname("..") -> "."
basename("/usr/lib") -> "lib"
basename("/usr/") -> "usr"
basename("usr") -> "usr"
basename("/") -> "/"
basename(".") -> "."
basename("..") -> ".."
The `lsdir()` function returns a sorted array containing the names
of all entries within the given directory path, without the common
"." and ".." entries. If the given path is not a directory, cannot
be opened or if another system level occurs, null is returned and
`fs.error()` can be used to query details.
The function takes an optional second argument which may be either a
regular expression value or a string. In case a regular expression
is given, each directory entry is matched against it. In case a
string is provided, it is treated as wildcard (glob) pattern and
only directory entries matching the pattern are considered.
Examples:
lsdir("/sys/class/net") -> [ "eth0", "lo", "wlan0" ]
lsdir("/proc", /^[0-9]+$/) -> [ "1", "4", "12", ... ]
lsdir("/sys/class/block/", "sd?3") -> [ "sde3", "sdf3" ]
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
|
Extract the wiphy index from the response message and use it to select
the item to merge the fragment information into.
Signed-off-by: Jo-Philipp Wich <jo@mein.io>
|
