Zeyad Yasser

I wasn't sure whether to post this here or at podman, I thought I should be able to do post-copy memory restore using config files without help from podman. Thanks...

Thanks for the quick response. I was not looking into live migrating the memory pages over TCP, just utilizing lazy pages for better restore response time until the pages are...

`sudo strace -fo /tmp/criu-lazy-strace.log -s 1024 criu lazy-pages -D . -o lazy-pages.log -vvvv` [criu-lazy-strace.log](https://github.com/checkpoint-restore/criu/files/8893522/criu-lazy-strace.log) @rppt [lazy-pages.log](https://github.com/checkpoint-restore/criu/files/8893523/lazy-pages.log) [core-dump](https://github.com/checkpoint-restore/criu/files/8893572/core-dump.log) @avagin /proc/sys/vm/unprivileged_userfaultfd was set to 0, I changed it to 1 but still...

[criu-lazy-strace-2.log](https://github.com/checkpoint-restore/criu/files/8893826/criu-lazy-strace-2.log) [lazy-pages-2.log](https://github.com/checkpoint-restore/criu/files/8893827/lazy-pages-2.log) ``` 17314 write(4, "(00.002180) uffd: Waiting for incoming connections on lazy-pages.socket\n", 72) = 72 17314 socket(AF_UNIX, SOCK_STREAM, 0) = 5 17314 unlink("lazy-pages.socket") = 0 17314 bind(5, {sa_family=AF_UNIX, sun_path="lazy-pages.socket"},...

It doesn't seem to have anything related to the container itself. Tested the flow on a generic container gave the same results. ```sudo podman run --runtime=runc -dt -p 8080:80/tcp docker.io/library/httpd```

`zdtm/static/env00` test passes. I applied the first patch, the problem unfortunately still persists. Linux: `Linux pc 5.17.13-200.fc35.x86_64 #1 SMP PREEMPT Mon Jun 6 14:38:57 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux`...

I noticed that if I changed `sendfd` to an fd other than the thread's `uffd`, the fd is sent successfully. I just tried sending fd 1. https://github.com/checkpoint-restore/criu/blob/criu-dev/criu/uffd.c#L224

> I was able to reproduce this issue on my host. Something is wrong with selinux rules. In ftrace output, I see this call trace: scm_detach_fds -> __receive_fd -> security_file_receive...

@sergiocazzolato could you please also extend the fix to the LXD backend? ~~It might be a good idea to extend the LXD spread test to run with other system variants...

@thp-canonical you could use the exclude config in spread.yaml to exclude any unwanted files (if their pattern is knowm), please check https://github.com/snapcore/spread/blob/master/README.md#including.