boinc_client gets blocked on XOpenDisplay () when 2 Xorg sessions are open

[paulez]

When two graphic sessions are opened on my system with only one active, boinc_client gets eventually stuck forever waiting on XOpenDisplay(). Killing the non-active Xorg session or switching back to it will unstuck boinc_client.

% ps aux|grep boinc
boinc      616  0.1  0.1  76512 16276 ?        SNsl oct.25  11:27 /usr/bin/boinc_client --dir /var/lib/boinc --redirectio
boinc     6469 27.0  2.7 375236 227700 ?       SNl  oct.28 471:12 ../../projects/boinc.bakerlab.org_rosetta/minirosetta_3.78_x86_64-pc-linux-gnu @G187193_1-50_MAP_hyb_TMP_v02_i01_t000__krypton.flags -in:file:boinc_wu_zip G187193_1-50_MAP_hyb_TMP_v02_i01_t000__krypton.zip -run:protocol jd2_scripting -silent_gz -mute all -out:file:silent default.out -nstruct 10000 -cpu_run_time 28800 -boinc:max_nstruct 600 -checkpoint_interval 120 -database minirosetta_database -in::file::zip minirosetta_database.zip -boinc::watchdog -run::rng mt19937 -constant_seed -jran 1240652
boinc     8431 15.0  0.0      0     0 ?        ZN   oct.28 238:57 [minirosetta_3.7] <defunct>
boinc     9481 18.7  0.0      0     0 ?        ZN   oct.28 267:20 [wcgrid_mcm1_7.3] <defunct>
boinc     9770  7.4  0.0      0     0 ?        ZN   oct.28 102:17 [minirosetta_3.7] <defunct>
boinc     9822 18.9  0.0      0     0 ?        ZN   oct.28 257:39 [wcgrid_mcm1_7.3] <defunct>
boinc    14127  0.0  0.0      0     0 ?        ZN   oct.28   0:06 [primegrid_genef] <defunct>
boinc    14128 17.8  0.0      0     0 ?        ZN   oct.28 226:39 [wcgrid_mcm1_7.3] <defunct>
paul     21201  0.2  1.4 2969360 120644 ?      Sl   19:10   0:00 /usr/bin/boincmgr
paul     21295  0.0  0.0  11204   896 pts/1    S+   19:12   0:00 grep --color=auto boinc

Backtrace (incomplete) but we can see boinc_client is stuck waiting on XOpenDisplay () to return.

Attaching to process 616
[New LWP 827]
[Thread debugging using libthread_db enabled]
Using host libthread_db library "/lib64/libthread_db.so.1".
0x00007f50e53a7a3d in poll () from /lib64/libc.so.6
(gdb) bt
#0  0x00007f50e53a7a3d in poll () from /lib64/libc.so.6
#1  0x00007f50e50aea9b in ?? () from /usr/lib64/libxcb.so.1
#2  0x00007f50e50ac693 in xcb_connect_to_fd () from /usr/lib64/libxcb.so.1
#3  0x00007f50e50b03d9 in xcb_connect_to_display_with_auth_info () from /usr/lib64/libxcb.so.1
#4  0x00007f50e63d674a in _XConnectXCB () from /usr/lib64/libX11.so.6
#5  0x00007f50e63c7702 in XOpenDisplay () from /usr/lib64/libX11.so.6
#6  0x0000000000494494 in ?? ()
#7  0x00000000004948cf in ?? ()
#8  0x000000000042090e in ?? ()
#9  0x00000000004732c0 in ?? ()
#10 0x00000000004084a0 in ?? ()
#11 0x00007f50e52e9640 in __libc_start_main () from /lib64/libc.so.6
#12 0x0000000000408a59 in ?? ()

I suspect that is because the non active Xorg server doesn't answer requests until it gets active again. As a result the boinc_client process gets fully stuck and unit processing eventually stops, as computing processes eventually get into zombie state when they finish their work.

[makeasnek]

A bounty has been started by The SCI to fix this issue, it is currently at $100 USD. You can contribute to, see the total amount of, and claim it at the link below. The bounty will continue to increase until it is claimed. TheSCInitiative/bounties#6