On 09/26/2018 05:25 PM, David C. Rankin wrote: > All, > > After recent updates, hylafax/sendfax fails with the error: > > sendfax: 460 Failed to submit job 110: Unable to open scheduler FIFO: No such > device or address. > > it had been working find up until the last week or two. Reverting to the LTS > kernel resolves the issue. There are no obvious errors in the journal. > > Does anyone know why sendfax would begin failing with "Unable to open > scheduler FIFO: No such device or address." with the current kernel, but work > fine on LTS? > > (I have confirmed this behavior on two separate Arch servers) > The last successful outbound fax before this failure began was 9/19. Since that time the only relevant updates have been: linux (4.18.8.arch1-1 -> 4.18.9.arch1-1) hylafax (6.0.6-8 -> 6.0.7-1) Since hylafax 6.0.7-1 works fine with LTS, that points to linux 4.18.9 being the problem. What would I check to determine why sendfax is unable to open scheduler FIFO (No such device or address) when running on linux 4.18.9? Permissions under /var/spool/hylafax are the same under both kernels: # l /var/spool/hylafax/ total 76 drwxr-xr-x 17 uucp uucp 4096 Sep 25 22:16 . drwxr-xr-x 10 root root 4096 Dec 29 2013 .. drwx------ 2 uucp uucp 4096 May 21 2013 archive drwxr-xr-x 2 uucp uucp 4096 Sep 25 22:16 bin drwxr-xr-x 2 uucp uucp 4096 Sep 26 15:49 client drwxr-xr-x 2 uucp uucp 4096 Sep 25 22:16 config drwxr-xr-x 2 uucp uucp 4096 Oct 23 2017 dev drwx------ 2 uucp uucp 4096 Sep 26 15:49 docq drwx------ 2 uucp uucp 4096 Sep 20 00:00 doneq drwxr-xr-x 3 uucp uucp 4096 Sep 26 00:00 etc drwxr-xr-x 2 uucp uucp 4096 Sep 19 15:12 info drwxr-xr-x 2 uucp uucp 4096 Sep 26 10:25 log drwx------ 2 uucp uucp 4096 May 21 2013 pollq drwxr-xr-x 2 uucp uucp 4096 Sep 26 10:25 recvq drwx------ 2 uucp uucp 4096 Sep 26 15:49 sendq drwxr-xr-x 2 uucp uucp 4096 Oct 23 2017 status drwx------ 2 uucp uucp 4096 Sep 26 15:49 tmp -r--r--r-- 1 root root 5426 Sep 25 07:29 COPYRIGHT prw------- 1 uucp uucp 0 Sep 25 07:29 FIFO prw------- 1 uucp uucp 0 Sep 20 19:57 FIFO.ttyS0 The FIFO is there owned by uucp:uucp in both cases. Did the latest kernel tighten some type of security that would effect mkfifo for hylafax in some way? This caught me somewhat off-guard as both Arch servers are used in a production capacity and I had a fax that needed to go to Scottsdale before close of business. Any thoughts welcomed. I'll also post to the hylafax list and report back if there is any information there. -- David C. Rankin, J.D.,P.E.
