Age | Commit message (Collapse) | Author |
|
Also change path.exists() and !path.isdir() to a single
path.isfile() which should be equal.
Enables running tests against image recipes which are not called
"bla-image" but plain "bla". Currently they fail with do_testimage/runqemu
error:
runqemu - ERROR - Unknown path arg /home/builder/src/base/build/tmp_qemuarm64/deploy/images/qemuarm64/img-qemuarm64.rootfs.wic
Suggested-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Mikko Rapeli <mikko.rapeli@linaro.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
When invoking runqemu with Python 3.12, the following warning is
encountered:
|SyntaxWarning: invalid escape sequence '\.'
This is because the interpreter scans the string before it is processed
by the regex module, and it interprets the backslash as part of an
escape sequence, but not a standard one. This will be registered as an
error rather than a warning in future Python versions. To avoid the it,
simply add an extra backslash so that Python doesn't misinterpret the
string, while the regex parser still sees an escaped '.' character.
Signed-off-by: Trevor Gamblin <tgamblin@baylibre.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Add support for qmp sockets and defaults to unix:qmp.sock if unspecified
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Eilís 'pidge' Ní Fhlannagáin <pidge@baylibre.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
When using a squashfs filesystem type, runqemu requires specifying the
full path to the image because it doesn't list squashfs types in its
fstypes variable. Add them to provide the same support as other
filesystem types.
Signed-off-by: Logan Gunthorpe <logang@deltatee.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
|
|
This disables Nagle's algorithm for our tcp serial connections which may
be causing data transfer issues.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
I have a theory that some of the console boot issues we're seeing are due to
starting images with three serial ports yet only starting gettys on two of them.
This means that occasionally, depending on the port numbering we may not get
a login prompt on the console we expect it on.
To fix this, change the runqemu code so that if serial ports are passed in on
the commandline (as is the case in automated testing), we don't add any other
GUI serial consoles.
We do need to make sure we do have at least two serial ports since we don't want
getty timeout warnings.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
qemu itself is not helpful when render nodes exist, but can't be opened:
qemu-system-x86_64: egl: render node init failed
To fix this, users likely need to
* modprobe vgem (presence when physical graphic card is absent or has a driver without
support for render nodes, such as many older cards found in server machines)
* add their user to "render" group to write to /dev/dri/renderD* (permissions)
With this change runqemu should print hints for the above as appropriate from probing the nodes.
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
|
|
With kernel 6.2 and later network devices are renamed by systemd. This does not
match with the current network device naming assumed in our configuration.
We may or may not change that naming but for now, pass the right kernel commandline
so things work as expected with newer kernels and removing a blocker on upgrading
to the 6.4 kernel by default.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
When runqemu tries to call runqemu-ifup to create tap devices, it checks the
output of runqemu-ifup to get the newly created tap device.
The behavior of runqemu-ifup was recently modified along with its output, it
no longer expects the uid parameter to be passed and it prints out a warning
if it was, since this warning was now part of the output runqemu tries to parse
it and convert it to an int() which proved impossible.
Pass the correct arguments to the runqemu-ifup call and echo the warning
to stderr instead to make sure its not being parsed and used by runqemu in any
case.
Signed-off-by: Alejandro Enedino Hernandez Samaniego <alejandro@enedino.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
logger.warn() has been deprecated, logger.warning() should be used instead.
Signed-off-by: Alejandro Enedino Hernandez Samaniego <alejandro@enedino.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Support the new environment variable OE_TAP_NAME.
Signed-off-by: Adrian Freihofer <adrian.freihofer@siemens.com>
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Since https://git.yoctoproject.org/poky/commit/?id=51063c1e6ac
we need to pass exactly 2 arguments to runqemu-ifup, otherwise
the script will return an error since bindir is no longer being
used.
However the call to runqemu-ifup from runqemu is still passing
bindir as an argument resulting in the error mentioned above,
remove the bindir argument to fix this issue.
[YOCTO #15150]
Signed-off-by: Alejandro Enedino Hernandez Samaniego <alejandro@enedino.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
with unrelated processes
There is already a neat check_free_port() function for finding an available port
atomically, so use that and make two additional tweaks:
- no need to allocate two separate ports; per unfsd documentation they can be the same
- move lockfile release until after unfsd has been shut down and the port(s) used has been freed
[YOCTO #15077]
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Default network configuration requires tun/tap module and while being
usable it conflicts with tap devices created by VPN clients sometimes
and requires root permissions to use . While it's possible to work
this around it's not always feasible if network is not required
Add nonetwork option which can be specified if the network connectivity is
not needed and SDL/serial is enough to communicate with the image.
Signed-off-by: Pavel Zhukov <pazhukov@suse.de>
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
|
|
* when searching for qemuboot.conf
* don't assume that IMAGE_LINK_NAME is always
<rootfs>-<machine> (with <rootfs>-<machine>.qemuboot.conf)
* runqemu: use IMAGE_LINK_NAME set by testimage.bbclass or query with bitbake -e
* testimage.bbclass was setting DEPLOY_DIR which I don't see used
anywhere else, so I assume it was supposed to be DEPLOY_DIR_IMAGE as mentioned
in corresponding runqemu code, do the same with IMAGE_LINK_NAME variable
* add virtual/kernel as bitbake -e target in run_bitbake_env to make
sure IMAGE_LINK_NAME is defined (kernel-artifact-names.bbclass inherits
image-artifact-names.bbclass as well)
* improve .qemuboot.conf search
1st search for file matching the rootfs and only when not found
try again with .rootfs suffix removed
[YOCTO #12937]
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
This reverts commit 82e67b82ea8e12aa0b7b9db1d84fec0436dec71b.
It was commited as part of https://bugzilla.yoctoproject.org/show_bug.cgi?id=12301
for kernels < 4.15, as of now oldest builder kernel we have is 4.15 on
ubuntu 18.04 so we should not require this workaround. Moreover, this
fixes an smp problem with qemux86 where no matter what -smp <x> option
is used, qemu always starts with single core.
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
|
|
The commands passed to subprocess are tuples which when passed to a %
format are treated as multiple format arguments instead of a single
argument to be coerced by "%s". This results in a TypeError being
raised with python claiming that not all arguments were consumed.
Fix this by wrapping the command tuple in a str() call, as is done for
the logging strings
[YOCTO #15078]
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
|
|
Signed-off-by: Jialing Zhang <zhangjialing@loongson.cn>
Signed-off-by: Qizheng Zhu <zhuqizheng@loongson.cn>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
|
|
host distro
With mesa 23.0, it is not longer possible to use the host drivers, as
mesa upstream has added strict checks for matching builds between
drivers and libraries that load them.
Add a check and a hint to runqemu so that there is a helpful error when
there is no native/nativesdk opengl/virgl support.
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
|
|
* to better indicate how it's used in get_first_file
* cmd* is used in other places for actual shell commands
to execute
* RunQemuError('KERNEL not found: %s, %s or %s' % cmds)
also looked weird to me, but that works (to my python-noob surprise)
[YOCTO #12937]
Signed-off-by: Martin Jansa <Martin.Jansa@gmail.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
|
|
Add support to the runqemu script for a new option, 'guestagent', that
enables the virtio serial port for host-to-guest communication.
Signed-off-by: Brenda Streiff <brenda.streiff@ni.com>
Signed-off-by: Clément Péron <peron.clem@gmail.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
This will produce a more useful hint for those setting
up or testing virgl headless:
runqemu - ERROR - No render nodes found in /dev/dri: ['by-path', 'card0']. If /dev/dri/renderD* is absent due to lack of suitable GPU, 'modprobe vgem' will create one suitable for mesa llvmpipe software renderer.
as qemu itself isn't helpful:
alex@Zen2:/srv/storage/alex/yocto/build-64-alt$ qemu-system-x86_64 -display egl-headless
qemu-system-x86_64: egl: no drm render node available
qemu-system-x86_64: egl: render node init failed
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
qemu doesn't always behave well and can hang too.
kill it with force if it was still alive. Move clean up
commands into cleanup() function.
Signed-off-by: Mikko Rapeli <mikko.rapeli@linaro.org>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
|
|
Signed-off-by: Xiaotian Wu <wuxiaotian@loongson.cn>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
This extra options to be appended to the nfs rootfs options in kernel
boot arg.
Example config with qemuppc64 in machine config:
add r/w size in the nfs rootfs extra option:
QB_NFSROOTFS_EXTRA_OPT = "wsize=524288,rsize=524288"
re-build and runqemu with nfs again, we can observe the kernel command
line added our defined value in QB_NFSROOTFS_EXTRA_OPT:
Kernel command line: root=/dev/nfs nfsroot=10.0.2.2:/home/xchen5/testing/build/tmp-glibc/deploy/images/qemuppc64/testnfs,nfsvers=3,port=3049,tcp,mountport=3048,wsize=524288,rsize=524288 rw mem=256M ip=dhcp console=hvc0 console=hvc0 nohugevmalloc
Signed-off-by: Xiangyu Chen <xiangyu.chen@windriver.com>
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
|
|
These enable running custom shell setup and cleanup commands
before and after qemu. Enables machine configurations to for
example run qemu with swtpm to emulate TPM devices.
Example config with meta-tpm2 based swtpm in machine config:
* image recipe depens on swtpm-native to get the native tools to PATH,
same handling as qemu itself
do_testimage[depends] += "swtpm-native:do_populate_sysroot"
* startup commands for swtpm daemon, note that swtpm
socket file has 107 character limit and absolute paths to build environment
will not work
QB_SETUP_CMD = " \
test -d '${IMAGE_BASENAME}_swtpm' || ( mkdir -p '${IMAGE_BASENAME}_swtpm' && \
swtpm_setup --tpmstate '${IMAGE_BASENAME}_swtpm' --tpm2 --pcr-banks sha256 ); \
swtpm socket --tpmstate dir='${IMAGE_BASENAME}_swtpm' \
--ctrl type=unixio,path='${IMAGE_BASENAME}_swtpm/swtpm-sock' \
--log level=40 --tpm2 -t -d \
"
* qemu startup command in machine config is configured enable swtpm device
QB_OPT_APPEND += "-chardev socket,id=chrtpm,path='${IMAGE_BASENAME}_swtpm/swtpm-sock' -tpmdev emulator,id=tpm0,chardev=chrtpm -device tpm-tis-device,tpmdev=tpm0"
* in this case, swtpm daemon stops automatically with qemu machine, but
QB_CLEANUP_CMD could be used to kill a specific process and wipe
temporary files
Now runqemu and testimage.bbclass can be used with swtpm.
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>
|
|
Rather, set it similarly to the overall network config.
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
|
|
With default slirp port forwarding config qemu listens on TCP ports
2222 and 2323 on all IP addresses available on the build host. Most
use cases with runqemu only need it for localhost and it is not
safe to run qemu images with root login without password enabled
and listening on all available, possibly Internet reachable network
interfaces. Limit qemu port forwarding to localhost 127.0.0.1 IP
address. Now qemu machine SSH and telnet ports are only
reachable from the build host machine, not full Internet.
If qemu machine needs to be reachable from network, then it can
be enabled via local.conf or machine config variable QB_SLIRP_OPT:
QB_SLIRP_OPT = "-netdev user,id=net0,hostfwd=tcp::2222-:22"
Signed-off-by: Mikko Rapeli <mikko.rapeli@linaro.org>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
|
|
Instead of changing the script environment to affect the child
processes, make a copy of the environment with modifications and pass
that to subprocess.
Specifically, when dri rendering is enabled, LD_PRELOAD was being passed
to all processes created by the script which resulted in other commands
(e.g. stty) exiting with a failure like:
/bin/sh: symbol lookup error: sysroots-uninative/x86_64-linux/lib/librt.so.1: undefined symbol: __libc_unwind_link_get, version GLIBC_PRIVATE
Making a copy of the environment fixes this because the LD_PRELOAD is
now only passed to qemu itself.
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
|
|
The code to parse arguments was inadvertently skipping all arguments in
the elif block after gl-es if it was specified on the command line.
Signed-off-by: Joshua Watt <JPEWhacker@gmail.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
|
|
In order to be able to debug issues on the host that have an impact on the
guest, it is useful to get the uptime of the host while starting so we can
match with the events in dmesg. Also include the uptime when cleaning up.
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Signed-off-by: Mark Hatle <mark.hatle@kernel.crashing.org>
Signed-off-by: Mark Hatle <mark.hatle@amd.com>
Signed-off-by: Luca Ceresoli <luca.ceresoli@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
runqemu auto generating the KERNEL_CMDLINE values and specifying using
-append option to qemu boot command which will lead to override the
kernel_cmdline/bootargs which are specified in DTB when using -dtb option.
Add new macro QB_KERNEL_CMDLINE to specify not to add the runqemu
generated KERNEL_CMDLINE values instead use which are in the DTB if
value defined as 'none'.
Add provision to override bootargs using # runqemu bootparams="root=/dev/ram0"
Signed-off-by: Raju Kumar Pothuraju <raju.kumar-pothuraju@xilinx.com>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The graphics option auto detection logic added in 7f78bb7a was not
checking if the publicvnc option had been specified, meaning that
it would be ignored and the auto detection result used instead.
Add setting a flag variable in the argument parsing and check
it along with the ones for the other graphics backend options.
Signed-off-by: Scott Murray <scott.murray@konsulko.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Running "runqemu qemux86 kvm" when qemu is configured for sdl and/or gtk
display output currently leads to a poor user experience with no cursor
and corrupted fonts in the gtk case. This is due to no options being
passed to qemu which leads to the loss of the font envirornment variable
and the show-cursor option.
If the user hasn't specified a display type, grep the output of
"qemu-system-xxx --help" for the display types and pick the "best"
which ensures our config is passed in. That resolves the gtk font issue
and the cursor issue with both sdl and gtk.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
When setting memory to 4GB, qemu is only running with 2GB for x86_64.
Avoid this by removing the mem= option to the kernel and letting the
qemu configuration handle it for x86 in a similar way to mips.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Among other things, fix misspellings of:
- absolute
- deprecated
- suitable
- handle
and a bunch of other things.
Signed-off-by: Robert P. J. Day <rpjday@crashcourse.ca>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Some of the host distributions build the drivers in a way (RPATH/RUNPATH)
that tricks uninative loader into loading pieces of the host libc, if
the same pieces haven't been previously loaded by native binaries. Mixing
the two libc versions leads to failures.
This change ensures that the correct (uninative) versions are always in use.
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The wrapper is executed by host bash, and host bash refuses to run
when pseudo libc is preloaded via LD_PRELOAD (which is the case
when gl is enabled).
Only the fontconfig setting is carried over as local testing showed that
only that is necessary for the gui to look ok nowadays; adjust further
if necessary.
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
New systemd is actually parsing this in systemd-network-generator
and fails if it is not fully formed. 'off' means 'static ip, do nothing':
https://www.kernel.org/doc/Documentation/filesystems/nfs/nfsroot.txt
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
If runqemu is killed, check that we have a valid PID for the qemu before
sending a kill() to it.
[ YOCTO #14651 ]
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Optionally allow to set QB_KERNEL_ROOT to e.g. "/dev/vda ro" to mount the
rootfs reay-only in Qemu.
Signed-off-by: Adrian Freihofer <adrian.freihofer@siemens.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Not all machine definitions need to have SERIAL_CONSOLES defined, but
runqemu currently will fail with the following script error if not
present:
Traceback (most recent call last):
File "/builder/meta-arm/poky/scripts/runqemu", line 1604, in main
config.setup_final()
File "/builder/meta-arm/poky/scripts/runqemu", line 1446, in setup_final
self.setup_serial()
File "/builder/meta-arm/poky/scripts/runqemu", line 1381, in setup_serial
self.kernel_cmdline_script += ' console=%s' %entry.split(';')[1]
IndexError: list index out of range
To get around this issue, add a sanity check to runqemu to avoid the
parsing of SERIAL_CONSOLES if empty.
Signed-off-by: Jon Mason <jdmason@kudzu.us>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Correct two issues:
1. Looking for dri.pc is only needed when gl is enabled.
2. virtio-vga-gl works only when gl is enabled via -display,
otherwise virtio-vga needs to be selected.
Signed-off-by: Alexander Kanavin <alex@linutronix.de>
Signed-off-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
|
|
Add the ability to replace DEPLOY_DIR_IMAGE with that path in
QB_OPT_APPEND. This allows for anything present in DEPLOY_DIR_IMAGE to
be passed into the qemu parameters. This is especially useful if you
want to run multiple flash images (as -bios only allows for one).
Signed-off-by: Jon Mason <jdmason@kudzu.us>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
qemu 6.1 has (perhaps unintentionally) disabled virgl in -vga virtio:
https://gitlab.com/qemu-project/qemu/-/issues/586
Signed-off-by: Alexander Kanavin <alex.kanavin@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
This is particularly useful for llvm-accelerated GL rendering from qemu guest to
an offscreen buffer (accessible over vnc or spice) using llvmpipe on the
host, rather than using unaccelerated swrast renderer in the guest.
This is the best that can be done in the absence of a host GPU with render node
support (such as old Matrox cards common in servers, or fully virtualized cloud
environments with no GPU at all).
Note: even though NVidia blob drivers do support render nodes, they do not
support gbm (yet?), and so rendering will fall back to llvmpipe as well even when
the system has a 3000 euro NVidia GPU. Cue Linus picture.
Signed-off-by: Alexander Kanavin <alex.kanavin@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|