Bug 186803
| Summary: | nscd cache make groupadd error in scripts | ||
|---|---|---|---|
| Product: | [Fedora] Fedora | Reporter: | Kazutoshi Morioka <morioka> |
| Component: | shadow-utils | Assignee: | Peter Vrabec <pvrabec> |
| Status: | CLOSED RAWHIDE | QA Contact: | David Lawrence <dkl> |
| Severity: | medium | Docs Contact: | |
| Priority: | medium | ||
| Version: | 5 | CC: | jakub, k.georgiou |
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | All | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | Bug Fix | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2006-04-05 11:45:51 UTC | Type: | --- |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
That's a bug in shadow-utils, it doesn't properly notify nscd that it needs
to flush its cache.
shadow-4.0.14/lib/nscd.c has:
int nscd_flush_cache (char *service)
{
int sock = nscd_open_socket ();
request_header req;
struct iovec iov[2];
ssize_t nbytes;
if (sock == -1)
return -1;
req.version = NSCD_VERSION;
req.type = INVALIDATE;
req.key_len = strlen (service);
iov[0].iov_base = &req;
iov[0].iov_len = sizeof (req);
iov[1].iov_base = service;
iov[1].iov_len = req.key_len;
nbytes = writev (sock, iov, 2);
close (sock);
return (nbytes != iov[0].iov_len + iov[1].iov_len ? (-1) : 0);
}
which is wrong, the INVALIDATE request's key is supposed to contain the trailing
'\0' character and req.key_len is supposed to include that as well.
Compare that to what nscd does for nscd -i:
if (strcmp (arg, "passwd") == 0)
req.key_len = sizeof "passwd";
else if (strcmp (arg, "group") == 0)
req.key_len = sizeof "group";
else if (strcmp (arg, "hosts") == 0)
req.key_len = sizeof "hosts";
else
return ARGP_ERR_UNKNOWN;
req.version = NSCD_VERSION;
req.type = INVALIDATE;
So, shadow-utils needs to change the
req.key_len = strlen (service);
line in lib/nscd.c to:
req.key_len = strlen (service) + 1;
fixed in FC5(shadow-utils-4.0.14-6.FC5) and devel |
Description of problem: While nscd is running, chgrp after groupadd fails with group does not exist. For example, preinstall script of screen does groupadd to make "screen" group, and then, /usr/bin/screen and /var/run/screen/ is created with the "screen". but this is sometimes fail, and screen do not work. Version-Release number of selected component (if applicable): # rpm -q glibc glibc-2.4-4 # rpm -q screen screen-4.0.2-12 How reproducible: sometimes Steps to Reproduce: 1. service nscd start 2. groupadd foogroup ; groupdel foogroup ; sleep 1 3. repeat 2 some times. Actual results: [root@myhost ~]# groupadd foogroup ; groupdel foogroup ; sleep 1 [root@myhost ~]# groupadd foogroup ; groupdel foogroup ; sleep 1 [root@myhost ~]# groupadd foogroup ; groupdel foogroup ; sleep 1 [root@myhost ~]# groupadd foogroup ; groupdel foogroup ; sleep 1 [root@myhost ~]# groupadd foogroup ; groupdel foogroup ; sleep 1 groupdel: group foogroup does not exist Expected results: No error reported. Additional info: This is another example of this problem: [root@myhost ~]# yum -y install screen Loading "installonlyn" plugin Setting up Remove Process snip Running Transaction Installing: screen [1/1]warning: group screen does not exist - using root Installing: screen ######################## [1/1]warning: group screen does not exist - using root Installing: screen ######################### [1/1] Installed: screen.i386 0:4.0.2-12 Complete! [root@myhost ~]# ls -ld /usr/bin/screen /var/run/screen -rwxr-xr-x 1 root root 346612 2æ 27 18:10 /usr/bin/screen drwxrwxr-x 2 root root 4096 2æ 27 18:10 /var/run/screen