Bug 8872 - Concurrent connections to sysvol fail in s3fs Configuration
Summary: Concurrent connections to sysvol fail in s3fs Configuration
Status: RESOLVED FIXED
Alias: None
Product: Samba 4.0
Classification: Unclassified
Component: DCE-RPCs and pipes (show other bugs)
Version: 4.0 alpha 18
Hardware: All All
: P5 normal (vote)
Target Milestone: ---
Assignee: Andrew Bartlett
QA Contact: samba4-qa@samba.org
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2012-04-17 17:09 UTC by Arvid Requate
Modified: 2012-05-10 07:48 UTC (History)
2 users (show)

See Also:

Attachments
smb.conf, level 10 logs and the test script (750.45 KB, application/x-gzip)
2012-04-17 17:11 UTC, Arvid Requate 		no flags Details
Test script for testenv (2.13 KB, text/plain)
2012-04-21 11:49 UTC, Andrew Bartlett 		no flags Details
Patch to use generate_random() not random() for unique ID values (1.04 KB, patch)
2012-04-25 08:14 UTC, Andrew Bartlett 		no flags Details
View All Add an attachment (proposed patch, testcase, etc.)

Note You need to log in before you can comment on or make changes to this bug.
Description Arvid Requate 2012-04-17 17:09:28 UTC 
Samba4 with s3fs configuration currently does has issues handling concurrent login attempts, e.g. against the sysvol share. A shell-script run of 10 backgrounded smbclient processes usually shows at least two NT_STATUS_IO_TIMEOUT failure situations. Adding the "-k no" swith increases the failure rate significantly, only the first smbclient connection succeeds in that case.

root@dc:/home/sernet# ./client-login.sh 
OK
OK
OK
OK
OK
OK
OK
OK
Connection to \\dc.samba.private\sysvol failed - NT_STATUS_IO_TIMEOUT
Connection to \\dc.samba.private\sysvol failed - NT_STATUS_IO_TIMEOUT
Again without trying kerberos:
root@dc:/home/sernet# OK
Connection to \\dc.samba.private\sysvol failed - NT_STATUS_IO_TIMEOUT
Connection to \\dc.samba.private\sysvol failed - NT_STATUS_IO_TIMEOUT
Connection to \\dc.samba.private\sysvol failed - NT_STATUS_IO_TIMEOUT
Connection to \\dc.samba.private\sysvol failed - NT_STATUS_IO_TIMEOUT
Connection to \\dc.samba.private\sysvol failed - NT_STATUS_IO_TIMEOUT
Connection to \\dc.samba.private\sysvol failed - NT_STATUS_IO_TIMEOUT
Connection to \\dc.samba.private\sysvol failed - NT_STATUS_IO_TIMEOUT
Connection to \\dc.samba.private\sysvol failed - NT_STATUS_IO_TIMEOUT
Connection to \\dc.samba.private\sysvol failed - NT_STATUS_IO_TIMEOUT
Comment 1 Arvid Requate 2012-04-17 17:11:55 UTC 
Created attachment 7460 [details]
smb.conf,  level 10 logs  and the test script

It looks a bit like windind having a problem answering concurrent wbc_sids_to_xids calls:

[2012/04/17 18:54:05.892046,  5, pid=2385, effective(0, 0), real(0, 0)] ../source4/libcli/wbclient/wbclient.c:72(wb
c_sids_to_xids_send)
  wbc_sids_to_xids called
[...]
[2012/04/17 18:54:15.899285, 10, pid=2385, effective(0, 0), real(0, 0)] ../source3/lib/events.c:216(run_events_poll)
  Running timed event "tevent_req_timedout" 0x8660490
[2012/04/17 18:54:15.899431,  5, pid=2385, effective(0, 0), real(0, 0)] ../source4/libcli/wbclient/wbclient.c:118(wbc_sids_to_xids_recv)
  wbc_sids_to_xids_recv called
[2012/04/17 18:54:15.899483,  1, pid=2385, effective(0, 0), real(0, 0)] ../source3/smbd/sesssetup.c:264(reply_sesssetup_and_X_spnego)
  Failed to generate session_info (user and group token) for session setup: NT_STATUS_IO_TIMEOUT
Comment 2 Andrew Bartlett 2012-04-21 11:49:01 UTC 
Created attachment 7472 [details]
Test script for testenv

this reproduces trivially with

SELFTEST_TESTENV=plugin_s4_dc make testenv

and the attached (modified) test script
Comment 3 Andrew Bartlett 2012-04-25 08:14:42 UTC 
Created attachment 7501 [details]
Patch to use generate_random() not random() for unique ID values

This fixes the issue, and is in autobuild.
Comment 4 Stefan Metzmacher 2012-04-25 09:36:57 UTC 
(In reply to comment #3)
> Created attachment 7501 [details]
> Patch to use generate_random() not random() for unique ID values
> 
> This fixes the issue, and is in autobuild.

I think we should always fill in

id.pid = getpid();
id.task_id = generate_random();
id.vnn = NONCLUSTER_VNN;
id.unique_id = generate_random();
Comment 5 Matthias Dieter Wallnöfer 2012-05-10 07:48:06 UTC 
f10c63810077a6759a9df4e9c653066f9f355d96 seems to have fixed this.