accumulo-notifications mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Christopher Tubbs (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (ACCUMULO-3224) Shell should use nanos for auth timeout
Date Sun, 12 Oct 2014 04:41:33 GMT

    [ https://issues.apache.org/jira/browse/ACCUMULO-3224?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14168523#comment-14168523
] 

Christopher Tubbs commented on ACCUMULO-3224:
---------------------------------------------

[~busbey], do you mean on old kernels that don't have CLOCK_MONOTONIC support? No, I'm not
concerned about that case. CLOCK_MONOTONIC is POSIX and (from what I understand) has been
in the kernel since well before the 2.6.32 kernel in CentOS 6. And, in any case, that fallback
on those very old kernels is no worse than what was there before this was patched. Do you
have concerns about this case?

> Shell should use nanos for auth timeout
> ---------------------------------------
>
>                 Key: ACCUMULO-3224
>                 URL: https://issues.apache.org/jira/browse/ACCUMULO-3224
>             Project: Accumulo
>          Issue Type: Improvement
>          Components: shell
>            Reporter: Christopher Tubbs
>            Assignee: Christopher Tubbs
>             Fix For: 1.5.3, 1.6.2, 1.7.0
>
>         Attachments: 0001-ACCUMULO-3224-Use-nanoTime-in-the-shell-s-auth-timeo.patch
>
>          Time Spent: 0.5h
>  Remaining Estimate: 0h
>
> I was looking at the code done for ACCUMULO-3221 and noticed that we're using the system
clock instead of the JDK's internal relative time, System.nanoTime(). This is a problem, because
any auth timeout that depends on the system clock can be easily bypassed by changing the system
time.
> We can also do the time conversion more reliably with {{TimeUnit}} to avoid the potential
arithmetic bug identified in ACCUMULO-3221.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message