The current code is not fit for purpose when handling large files via ssh. In the strace
ptest case, we can end up with a 1.4GB archive being transferred for which every
byte is printed into the task logfile twice over. This is then sent over bitbake IPC
which compounds the problems.
Make the following improvements:
* when the output is large (over 64kb), don't print it
* use a bytearray for better concat performance since strings are slow for this
* when there is no ssh output, say that
* print periodic size status output rather than the data itself since this could be binary and/or large
* fix the killed process message logic which appeared broken
(From OE-Core rev: 57673a71b20e2bbb53e7652a709bdcb32c429b6b)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
(cherry picked from commit e7dd009a17dc902852983a82bce41bf78bb1e242)
Signed-off-by: Steve Sakoman <steve@sakoman.com>
The ssh target is currently well tailored to easily retrieve textual output
from a command run on a remote target. It could also be used to retrieve
raw data from a command run onto a remote target (for example, to feed this
data directly to another program), but it currently suffers two minor
issues preventing such use case:
- stderr is piped to stdout, so any error log will be mixed in the program
output
- the final output is decoded as utf-8 and stripped
Allow to return the raw, unmodified output by adding an optional "raw"
parameter. Keep it to False by default to preserve the current behavior.
When enabled, do not return a string but the raw output as bytes.
(From OE-Core rev: 8d05dc6e2284b7ed7c32a8215b9c8bf6f7dabf00)
Signed-off-by: Alexis Lothoré <alexis.lothore@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The target_dumper code is basically broken. It has been reading binary files
over the text base serial communication and runs at every command failure which
makes no sense. Each run might overwrite files from the previous run and the
output appears corrupted due to confusion from the binary data.
For now, remove the commands and the target dumper code as the command
and execution point are problematic. Also remove the same pieces of the monitor
code but leave the command list since in theory this can be moved to a more
useful place in the code.
(From OE-Core rev: a24d787987dccc95fdd95b7e85bf525a1c55b285)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The current code in ssh.py is terminating the ssh process that
does not finish its computation in a given timeout (when timeout
is passed), the SSHCall function is returning the process error
code.
The Openssl ssh before version 8.6_p1 is returning 0 when it is
terminated, from commit 8a9520836e71830f4fccca066dba73fea3d16bda
onwards (version >= 8.6_p1) ssh is returning 255 instead.
So for version of ssh older than 8.6_p1 when the SSHCall time out,
the return code will be 0, meaning success, which is wrong.
Fix this issue checking if the process has timeout (hence it's been
terminated) and checking if the returned code is 0, in that case
set it to 255 to advertise that an error occurred.
Add a test case excercising the timeout in the SSHTest, test_ssh
test function.
(From OE-Core rev: 948fecca1db4c7a30fcca5fcf5eef95cd12efb00)
Signed-off-by: Luca Fancellu <luca.fancellu@arm.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
It looks like there were further cases where orphaned processes may be left
behind since the .kill() calls may be unsuccessful if the process terminated
due to the terminate or through normal exit. In that situation .wait()
wouldn't have been called.
Further tweak the exit code paths to ensure .wait() is called to update the
returncode value before returning in all cases.
(From OE-Core rev: 0a0a1731e38edfa72a141e8fd8f2de52be562e94)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
As spotted by Joshua Watt, the returncode isn't set until .poll() or .wait()
is called so we need to call this after the .kill() call.
This fixes return code reporting so that timeouts for example now return an
exit code when they didn't before.
(From OE-Core rev: 3924e94214b5135369be2551d54fb92097d35e95)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
We have a suspicion that the read() call may return EAGAIN on the non-blocking
fd and this may truncate test output leading to some of our intermittent failures.
Tweak the code to avoid this potential issue.
(From OE-Core rev: a8920c105725431e989cceb616bd04eaa52127ec)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When qemu machine hangs, the ssh commands done by tests
are not timing out. do_testimage() task has last logs like this:
DEBUG: time: 1673531086.3155053, endtime: 1673531686.315502
The test process is stuck for hours, or for ever if the
executing command or test case did not set a timeout correctly.
The default 300 second timeout is not working when target hangs.
Note that timeout is really a "inactive timeout" since data returned
by the process will reset the timeout.
Make the process stdout non-blocking so read() will always return
right away using os.set_blocking() available in python 3.5 and later.
Then change from python codec reader to plain read() and make
the ssh subprocess stdout non-blocking. Even with select()
making sure the file had input to be read, the codec reader was
trying to find more stuff and blocking for ever when process hangs.
While at it, add a small timeout to read data in larger chunks if
possible. This avoids reading data one or few characters at a time
and makes the debug logs more readable.
close() the stdout file in all cases after read loop is complete.
Then make sure to wait or kill the ssh subprocess in all cases.
Just reading the output stream and receiving EOF there does not mean
that the process exited, and wait() needs a timeout if the process
is hanging. In the end kill the process and return the return value
and captured output utf-8 encoded, just like before these changes.
This fixes ssh run() related deadlocks when a qemu target hangs
completely.
(From OE-Core rev: 9c63970fce3a3d6029745252a6ec2bf9b9da862d)
Signed-off-by: Mikko Rapeli <mikko.rapeli@linaro.org>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Configure ssh client to test that connection with server is up.
If the server does not respond within a minute then the connection,
target machine or sshd daemon are stuck and it's better to exit
the command execution with errors.
Some tests can execute a long time without returning stdout/stderror
data and it's difficult to adjust timers for those cases if
connection to target machine or the target machine itself hangs
and output is not expected in minutes or even hours.
(From OE-Core rev: ba68ff04c5786eca7cd8dd44056705867dea8ac4)
Signed-off-by: Mikko Rapeli <mikko.rapeli@linaro.org>
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The output from is garbled otherwise and it's not
easy to remove debug output form real command output on target.
(From OE-Core rev: 917a70cbc43ac1c70c477b220c4115735457ef04)
Signed-off-by: Mikko Rapeli <mikko.rapeli@linaro.org>
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The underlying _run() method has an ignore_status argument and can
either return the exit code or raise an exception if it is non-zero.
Add the same argument to the run() method so that test cases can change
this value. It currently defaults to True as that is the existing
behaviour, but a follow-up patch will change this to False as test cases
should fail on errors unless told otherwise.
(From OE-Core rev: e244228730178d15a066a1428956de328cc09671)
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
The target_dumper property is only set by the QemuTarget subclass, so
assign a default value and check it isn't None before calling it.
(From OE-Core rev: add36cf85380bf7dc5339a9d65f951923317440b)
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This adds support for the Qemu Machine Protocol [0] extending
the current dump process for Host and Target. The commands are
added in the testimage.bbclass.
Currently, we setup qemu to stall until qmp gets connected and
sends the initialization and continue commands, this works
correctly. If the UNIX Socket does not exist, we wait an timeout
to ensure to socket file is created.
With this version, the monitor_dumper is created in OEQemuTarget
but then set in OESSHTarget as that's where we get the SSH failure
happens. Python's @property is used to create a setter/getter type
of setup in OESSHTarget to get overridden by OEQemuTarget.
By default the data is currently dumped to files for each command in
TMPDIR/log/runtime-hostdump/<date>_qmp/unknown_<seq>_qemu_monitor as
this is the naming convenstion in the dump.py code.
We use the qmp.py from qemu, which needs to get installed in the
recipe-sysroot-native of the target image.
[0] https://github.com/qemu/qemu/blob/master/docs/interop/qmp-spec.txt
(From OE-Core rev: 42af4cd2df72fc8ed9deb3fde4312909842fcf91)
Signed-off-by: Saul Wold <saul.wold@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This adds the dump_target support when the ssh command fails with
a 'No route to host'. This is will provide additional data when a
Qemu target fails to respond during autobuilder testing. This does
not fix 14002 [0], but may help track down why qemu looses networking
[0] https://bugzilla.yoctoproject.org/show_bug.cgi?id=14002
(From OE-Core rev: cef1a2b03b359c018911abc29db1895d0f46814c)
Signed-off-by: Saul Wold <saul.wold@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
[YOCTO #13802]
Make the scp failure non-fatal so the ltp tests continue to run and
the rest of the logs will be available to see afterwards.
(From OE-Core rev: 0f7d093038274f4f21f6cca39a96aac4f6c32ee3)
Signed-off-by: Mingde (Matthew) Zeng <matthew.zeng@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
start_new_session was added to python3 subprocess in v3.2 and
is meant to take the place of the common use of preexec_fn to
call os.setsid() in the child - as done here.
Update to use the new equivalent.
(From OE-Core rev: c3bd502f31a621394a2f6ea3ae90443ff4ac98a4)
Signed-off-by: André Draszik <git@andred.net>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Allow tests to access the listening port as just introduced.
Note that when using qemu this infrastructure shouldn't be
needed, but we still need to set the port to 0 so that
a listening port is determined automatically (e.g. by the
python http server).
(From OE-Core rev: ca58559fc30842c123907fb5cc68b3356c530862)
Signed-off-by: André Draszik <git@andred.net>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
We continued to see encoding problems with ssh commands run in oeqa. After much
research the conclusion was we should use ignore the errors since some occasional bad
locale encoding is better than the unicode decoding issues we were seeing which crashed
large parts of tests.
(From OE-Core rev: b3bae8aba5b897c1afcc8bf4ad8929251812d2b5)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This adds SPDX license headers in place of the wide assortment of things
currently in our script headers. We default to GPL-2.0-only except for the
oeqa code where it was clearly submitted and marked as MIT on the most part
or some scripts which had the "or later" GPL versioning.
The patch also drops other obsolete bits of file headers where they were
encoountered such as editor modelines, obsolete maintainer information or
the phrase "All rights reserved" which is now obsolete and not required in
copyright headers (in this case its actually confusing for licensing as all
rights were not reserved).
More work is needed for OE-Core but this takes care of the bulk of the scripts
and meta/lib directories.
The top level LICENSE files are tweaked to match the new structure and the
SPDX naming.
(From OE-Core rev: f8c9c511b5f1b7dbd45b77f345cb6c048ae6763e)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This code really needs to be rewritten to not split potential
multibyte characters, for now work around it to avoid exceptions like:
File "/home/pokybuild/yocto-worker/qa-extras2/build/meta/lib/oeqa/core/target/ssh.py", line 211, in run
data = reader.read(1024, 4096)
File "/usr/lib64/python3.6/codecs.py", line 503, in read
newchars, decodedbytes = self.decode(data, self.errors)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0x92 in position 0: invalid start byte
(From OE-Core rev: 17e87510378f2729208a8262695f28e1efe5eb4c)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
When running testimage task for core-image-sato-sdk, the following
error appeared.
UnicodeDecodeError: 'utf-8' codec can't decode byte 0x82 at position 0: invalid start byte
Checking the codes, I found it's caused by setting a 1024 limit for the
read method of the StreamReader object.
Comments from the manual:
"""
The chars argument indicates the number of decoded code points or bytes to
return. The read() method will never return more data than requested, but
it might return less, if there is not enough available.
"""
When running `systemctl status --full' on target, this error occurs.
This patch increase the bytes limit to 4096 to fix the error.
(From OE-Core rev: f1fad60ae3be4450aca6058d5665fb10a9148b44)
Signed-off-by: Chen Qi <Qi.Chen@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This ensures the console is kept clear of confusing output but that
the main logs contain good debugging information.
(From OE-Core rev: caeb5dcfbd3c1d71f8e0eb78b3dd45d5ce349d25)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
This can avoid UnicodeDecodeError error.
(From OE-Core rev: baa78420d8d8e716935852c9c7b749af0161395a)
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
There might be wild strings when read from target (especially when
reading ptest results), replace the errors to avoid breaking the test.
Fixed: (Not always happen)
$ bitbake core-image-sato -ctestimage
[snip]
status, output = self.target.run('ptest-runner', 0)
File "/buildarea/lyang1/poky/meta/lib/oeqa/core/target/ssh.py", line 84, in run
status, output = self._run(sshCmd, processTimeout, True)
File "/buildarea/lyang1/poky/meta/lib/oeqa/core/target/ssh.py", line 55, in _run
status, output = SSHCall(command, self.logger, timeout)
File "/buildarea/lyang1/poky/meta/lib/oeqa/core/target/ssh.py", line 258, in SSHCall
run()
File "/buildarea/lyang1/poky/meta/lib/oeqa/core/target/ssh.py", line 236, in run
output = process.communicate()[0].decode("utf-8")
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xc3 in position 4906: invalid continuation byte
[YOCTO #11547]
(From OE-Core rev: d0d2f892f0bed6adb5ec6fb59d64efcc97c83e19)
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
With this commit now it is possible to add targets with SSH for testing.
Most of it was imported for existing code, with improvements in log
handling.
[YOCTO #10234]
(From OE-Core rev: 3bc13548df4adb85f09467d200530a9c9f60da04)
Signed-off-by: Mariano Lopez <mariano.lopez@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
