[yoyoma2]

Here's a fix to this file for dd-wrt's chronic cron issues. This should help with a number of bugs such as 6744, 1955, 6683, 6688 and others.

One of the comments in defect 1955 states: "Not reproducable, works fine here and the code for it is correct". A quick look at database.c will show that "*system*" is repeatedly used as "unique key" for multiple distinct records. The fix simply makes the unique key actually unique by appending the filename to it and everything works.

Cron is actually always running fine every minute, but with the database unique keys being invalid, when changes occur to cron table files, anything can happen. To reproduce, touching /tmp/crontab and touching/editing/adding/removing files in /etc/cron.d/ should result in erratic cron behavior quick enough.

Fix it any way you like, but please don't sweep it under the now 14 year old rug.

[egc]

@yoyoma2, like I already posted, your patch looked valid to me and I contacted BS and he has already patched the source code. see: https://svn.dd-wrt.com/changeset/41189

So in the next build it should be good.

If you can test the next build and report your findings (you can PM me or report in the original thread) that would be great.

Thanks for your work, much appreciated

[giles02134]

That's great news! I'll wait for reports on the effectiveness of the patch in the coming release, before I update my own installation.

Thanks to all for following this through.

[yoyoma2]

Wow thats a quick turnaround!

Unfortunately I only have one router (Asus RT-AC56U: DD-WRT v3.0-r39960M kongac) with lots of stuff configured. To switch over to a BS build with minimal downtime I just ordered a cheap Netgear R6300v2 (almost same specs as RT-AC56U) from China but the sea shipping can take months. To be safe I wouldn't switch Kong->BS without a full reset.

I'm quite confident dd-wrt's cron reliability will greatly improve after 41189.

Thanks egc and BS for the quick reaction.

[egc]

[giles02134]

I did a firmware update yesterday (9-Oct) with Firefox 69.0.2 x64). The flash proceeded normally, and the router rebooted without a problem.

I can report that the cron issue has indeed been fixed. Before flashing, I removed all startup commands which referenced yoyoma2's cronwd.sh, and also removed the cronwd.sh cron job Administration->management->additional cron jobs. The router rebooted on its own at 3:00a.m. as specified in Administration->Keepalive.

Thanks to yoyoma2 for not only submitting a fix for the bug, but for the workaround he'd provided prior to the new release.

[kernel-panic69]

Well, looking at this while merging cron 4.1 to another project, it is obvious that DD-WRT doesn't use cron 4.1, and THEREIN probably lies the obvious problem. This patch maybe not necessary if DD-WRT was using cron 4.1 (the latest ever released...!) SMDH.

[yoyoma2]

The bug was not in the original Vixie cron code but rather in the integration changes made to fit dd-wrt's special needs (single user with multiple cron tables).

I'm all for using updated code versions but I don't think any of the subtle bug fixes would be noticed by dd-wrt users given this severely broken cron sat there for 14 years. BS's time is so limited he'll probably aim for changes that offer more bang for the buck.

This patch seems to fix dd-wrt's chronic cron issues as we see here.

[kernel-panic69]

Vixie CRON (ISC Cron, as it ended up with the last release at version 4.1) was released in 2004: https://ftp.isc.org/isc/cron/ ... I literally did a side-by-side diff of the entire source tree, and the version in DD-WRT is still a v3.x version, so it's older than 14 years. So, the likelihood of the bug still being there in 4.1 is unlikely. BUT, without the last official version of Vixie / ISC cron being merged, I can only speculate... but I guess if it now works, it now works.

[yoyoma2]

The patched database.c says "Copyright ... 1994" so yes it's older than 14 years so an update from BS would be nice. It also says "mark your changes (don't get me blamed for your possible bugs)". This was not done 14 years ago when this bug was added by BS.

Specifically in load_database(), an entire loop was added by BS reading any number of cron job files with arbitrary filenames. Only a single system cron job file and one cron job file per user (named after them) are assumed by the underlying code.

The fix in process_crontab() specifically accomodates the BS addition above so no, the 4.1 release most certainly does not accomodate for BS's 14 year old addition.

[kernel-panic69]

I guess I don't understand the purpose of hacking an older version of cron vs. updating to the last known version that seems to work fine without enhancement. But hey, good job. Thanks for fixing his likely unnecessary non-industry-standard-following hack, if that is what it is. I would have to extract the cron tree from all of my GPL stuff here to look and see. I was only saying that the code itself, especially in that file, is not line-for-line with upstream latest release of v4.1, feel free to check it for yourself. There is a lot of hacked old code in this firmware that I have noticed that was not updated to the latest version or has not been. Nothing new.

[yoyoma2]

[kernel-panic69]

So, you're trying to tell me that the cron code in DD-WRT is 4.1 or started as 4.1? Ok, yeah, let me wave the bullshit flag on that one. But again, thank you for fixing a bad hack.

[yoyoma2]

[kernel-panic69]