Message330407
> What is also useful to know, is that I'm observing this on a legacy RHEL 6 system *with a customized kernel* part of the Scyld ClusterWare (https://www.penguincomputing.com/products/software/scyld-clusterware/) that *cannot* be updated:
Do you mean this trouble is happened only on the system?
Or can this trouble be reproducible on normal Linux kernel?
---
I think there are no bug in Python. See this part of your
strace
```
pipe([5, 6]) = 0
fcntl(5, F_GETFD) = 0
fcntl(5, F_SETFD, FD_CLOEXEC) = 0
fcntl(6, F_GETFD) = 0
fcntl(6, F_SETFD, FD_CLOEXEC) = 0
clone(child_stack=0,
flags=CLONE_CHILD_CLEARTID|CLONE_CHILD_SETTID|SIGCHLD,
child_tidptr=0x2ad3c972adf0) = 239074
close(6) = 0
mmap(NULL, 1052672, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x2ad3ca8ed000
read(5,
```
Python creates pipe [5, 6], and set FD_CLOEXEC flag to them.
When `exec` in child process succeeded, these FDs must be closed by Linux kernel.
Then, `read(5,` in parent process must not be blocked. It must return 0 soon.
So I think there are some bug in your kernel, relating to CLOEXEC.
I am not interested in this issue much because 2.6.32 is too old to me and
I never used "Scyld ClusterWare".
Maybe, you can consult with the company. |
|
Date |
User |
Action |
Args |
2018-11-26 09:57:36 | methane | set | recipients:
+ methane, Henrik Bengtsson |
2018-11-26 09:57:36 | methane | set | messageid: <1543226256.83.0.788709270274.issue35305@psf.upfronthosting.co.za> |
2018-11-26 09:57:36 | methane | link | issue35305 messages |
2018-11-26 09:57:36 | methane | create | |
|