The test fails when running ``make test TARGETS="//test/syscalls:io_uring_setup_test_native"
1 reply
external/com_google_absl/absl/synchronization/internal/graphcycles.cc: In member function 'void absl::synchronization_internal::GraphCycles::RemoveNode(void*)':
external/com_google_absl/absl/synchronization/internal/graphcycles.cc:450:26: error: 'numeric_limits' is not a member of 'std'
450 | if (x->version == std::numeric_limits<uint32_t>::max()) {
| ^~~~~~~~~~~~~~
external/com_google_absl/absl/synchronization/internal/graphcycles.cc:450:49: error: expected primary-expression before '>' token
450 | if (x->version == std::numeric_limits<uint32_t>::max()) {
| ^
external/com_google_absl/absl/synchronization/internal/graphcycles.cc:450:52: error: '::max' has not been declared; did you mean 'std::max'?
450 | if (x->version == std::numeric_limits<uint32_t>::max()) {
| ^~~
| std::max
In file included from /usr/lib/gcc/x86_64-pc-linux-gnu/11.1.0/../../../../include/c++/11.1.0/algorithm:62,
from external/com_google_absl/absl/synchronization/internal/graphcycles.cc:38:
/usr/lib/gcc/x86_64-pc-linux-gnu/11.1.0/../../../../include/c++/11.1.0/bits/stl_algo.h:3467:5: note: 'std::max' declared here
3467 | max(initializer_list<_Tp> __l, _Compare __comp)
| ^~~
Target //test/syscalls:io_uring_setup_test_native failed to build
Use --verbose_failures to see the command lines of failed build steps.
And I've read here that
Just note that they require specific configuration to work. This is handled automatically by the test scripts in the scripts directory and they can be used to run tests locally on your machine. but I can't find the scripts directory anywhere...
That blurb refers to "For the above noted cases, the relevant runtime must be installed via runsc install before running."
Which is only the case for integration, image, and root tests IIUC
Ok. I will look into what the Makefile is doing then. It is strange that the tests and its dependencies compiles fine in the docker but not using bazel directly
That error looks like it would be fixed by https://github.com/abseil/abseil-cpp/commit/5bf048b8425cc0a342e4647932de19e25ffd6ad7#diff-c2c26d26953e79ce20dd00dccad9e10e0249a17861480d58e73c4616786f2679, which is newer than our version of Abseil
Skallwar (Esteban Blanc): you could try updating https://cs.opensource.google/gvisor/gvisor/+/master:WORKSPACE;l=1275-1283 to the latest HEAD version of https://github.com/abseil/abseil-cpp.
1 reply
That's what I did in WORKSPACE:
# System Call test dependencies.
http_archive(
name = "com_google_absl",
sha256 = "1764491a199eb9325b177126547f03d244f86b4ff28f16f206c7b3e7e4f777ec",
strip_prefix = "abseil-cpp-278e0a071885a22dcd2fd1b5576cc44757299343",
urls = [
"https://github.com/abseil/abseil-cpp/archive/278e0a071885a22dcd2fd1b5576cc44757299343.tar.gz"
],
)
I used the latest release (7 days ago)
And remove
~/.cache/bazel to make sure it will be downloaded
Perhaps double check that
$(bazel info output_base)/external/com_google_absl/absl/synchronization/internal/graphcycles.cc contains the <limits> include?
1 reply
I'm new to bazel, is there something to do to "setup" the "workspace"?
No, changing the WORKSPACE should have been sufficient, even without clearing cache.
Let me give it a try
Update gRPC to
http_archive(
name = "com_github_grpc_grpc",
sha256 = "abd9e52c69000f2c051761cfa1f12d52d8b7647b6c66828a91d462e796f2aede",
strip_prefix = "grpc-1.38.0",
urls = [
"https://github.com/grpc/grpc/archive/v1.38.0.tar.gz",
],
)
The
grpc_deps() adds an absl dependency which takes precedence over ours since it comes first.
1 reply
Ah, makes sense. I can make a PR with these 2 changes if you want
Right, if the same dep
name is added twice, only the first is used. And grpc adds com_google_absl in https://github.com/grpc/grpc/blob/master/bazel/grpc_deps.bzl#L263. (I wish Bazel made multiple definition an error).
A PR would be great. IMO we should probably move our absl above gRPC so it actually does something.
2 replies
Now the test is correctly compiled but it fails anyway with
io_uring_setup returning -1 when using the native runner. How can it fails in the test but not as a standalone executable... Anyway thanks for the help
I'm running into a problem. I'm trying to debug runsc following this tutorial. But enven with
bazel build -c dbg //runsc:runsc, dvl is warning me Warning: debugging optimized function and I cannot print any arguments nor local variables:(dlv) args
t = ("*gvisor.dev/gvisor/pkg/sentry/kernel.Task")(0xc0001ee000)
args = gvisor.dev/gvisor/pkg/sentry/arch.SyscallArguments [...]
~r2 = (unreadable empty OP stack)
~r3 = (unreadable empty OP stack)
~r4 = (unreadable empty OP stack)
Any idea ?
The debugging instructions are missing a step of actually installing the binary as a docker runtime.
https://gvisor.dev/docs/user_guide/debugging/#debugger
https://gvisor.dev/docs/user_guide/debugging/#debugger
Hello, not sure if this is the right place... or minikube issue... I am following these instructions:
https://github.com/kubernetes/minikube/tree/master/deploy/addons/gvisor
and trying to get a simple app running using gVisor and it's stuck in
https://github.com/kubernetes/minikube/tree/master/deploy/addons/gvisor
and trying to get a simple app running using gVisor and it's stuck in
ContainerCreating status... regardless of the runtime class I use...? As soon as turn on the gVisor plugin i can't seem to create workloads successfully---
apiVersion: v1
kind: Namespace
metadata:
name: bbgv
spec: {}
---
apiVersion: apps/v1
kind: Deployment
metadata:
name: bbapp
namespace: bbgv
spec:
replicas: 1
selector:
matchLabels:
name: bbgvtest
template:
metadata:
labels:
name: bbgvtest
spec:
runtimeClassName: gvisor
containers:
- name: bbapp
image: busybox
args:
- /bin/sh
- -c
- >
echo "starting busybox"
i=0;
while true;
do
echo "$i: $(date) message from busybox gvisor";
echo "$(date) INFO $i message from busybox gvisor";
echo ""
i=$((i+1));
sleep 20;
doneSomewhat related to what I am trying to get to debugging... does gVisor allow inotify system calls, specifically inotify_init*() and inotify_add_watch()? is there anything needed to get that working? fluentbit tail plugin uses those to watch and tail log files... and from what I can tell it never gets any notifications...
[2021/06/11 03:09:38] [debug] [input:tail:tail.0] inotify watch fd=22
[2021/06/11 03:09:38] [debug] [input:tail:tail.0] scanning path /applogs/*.log
[2021/06/11 03:09:38] [debug] [input:tail:tail.0] inode=97 with offset=219 appended as /applogs/0.log
[2021/06/11 03:09:38] [debug] [input:tail:tail.0] scan_glob add(): /applogs/0.log, inode 97
[2021/06/11 03:09:38] [debug] [input:tail:tail.0] 1 new files found on path '/applogs/*.log'
[2021/06/11 03:09:38] [debug] [stdout:stdout.0] created event channels: read=24 write=25
[2021/06/11 03:09:38] [debug] [router] match rule tail.0:stdout.0
[2021/06/11 03:09:38] [ info] [sp] stream processor started
[2021/06/11 03:09:38] [debug] [input:tail:tail.0] inode=97 file=/applogs/0.log promote to TAIL_EVENT
[2021/06/11 03:09:38] [ info] [input:tail:tail.0] inotify_fs_add(): inode=97 watch_fd=1 name=/applogs/0.loglooks like things get setup ok, but it never seems to get any data as it's written to the log file.
You can break on something like
gvisor.dev/gvisor/pkg/sentry/socket/netstack.(*SocketOperations).Accept and then curl http://localost:8080/ to get it to hit the breakpoint.
If you debug code in runsc you should be able to see the code and print args and locals
1 reply
@bojand inotify is supported but you only get notificatations for changes that occur inside the sandbox. Any changes to files outside the sandbox won't generate events. Syscall support is summarized here:
https://gvisor.dev/docs/user_guide/compatibility/linux/amd64/
https://gvisor.dev/docs/user_guide/compatibility/linux/amd64/
I tried to used gdb but I have the same behavior:
(gdb) p params
No symbol "params" in current context.
huh, interesting, that may be a bug. Could you file an issue to attach the runsc debug+strace logs (https://gvisor.dev/docs/user_guide/debugging/)? The .boot and .gofer logs should be most relevant.
Sorry for the delay, added an issue google/gvisor#6180 , if somebody has a clue about what is causing it, I would highly appreciate it :)
@skallwar:matrix.org Can you create an issue explaining the problem and we can maybe follow up at some point. Given it's a pretty small scope, it probably won't be high priority though. If you wanted to do some more digging to figure out the problem that would be appreciated.
1 reply