swaylock hangs on acquiring fontconfig lock, preventing the session from being unlocked

[danielmussell]

swaylock 1.8.0/fontconfig 2.15.0 on Fedora 41.

I have this intermittent issue where after entering my password, swaylock hangs and cannot unlock the session. When this happens, I need to switch to another VT and kill the session, which is inconvenient to say the least. A backtrace from gdb gives

#0  futex_wait (futex_word=0x7f2d600a2bb0, expected=2, private=0) at ../sysdeps/nptl/futex-internal.h:146
#1  __GI___lll_lock_wait (futex=futex@entry=0x7f2d600a2bb0, private=0) at lowlevellock.c:49
#2  0x00007f2d68911d81 in lll_mutex_lock_optimized (mutex=0x7f2d600a2bb0) at pthread_mutex_lock.c:48
#3  ___pthread_mutex_lock (mutex=0x7f2d600a2bb0) at pthread_mutex_lock.c:93
#4  0x00007f2d6870e88f in FcMutexLock (m=<optimized out>) at /usr/src/debug/fontconfig-2.15.0-8.fc41.x86_64/src/fcmutex.h:122
#5  0x00007f2d6870ff65 in FcCacheFindByAddr (object=0x7f2d67aba1c0) at /usr/src/debug/fontconfig-2.15.0-8.fc41.x86_64/src/fccache.c:657
#6  FcCacheObjectReference (object=0x7f2d67aba1c0) at /usr/src/debug/fontconfig-2.15.0-8.fc41.x86_64/src/fccache.c:747
#7  IA__FcCharSetCopy (src=0x7f2d67aba1c0) at /usr/src/debug/fontconfig-2.15.0-8.fc41.x86_64/src/fccharset.c:369
#8  IA__FcCharSetCopy (src=0x7f2d67aba1c0) at /usr/src/debug/fontconfig-2.15.0-8.fc41.x86_64/src/fccharset.c:362
#9  0x00007f2d68726cb8 in IA__FcValueSave (v=...) at /usr/src/debug/fontconfig-2.15.0-8.fc41.x86_64/src/fcpat.c:113
#10 0x00007f2d6872fd1f in FcFontSetMatchInternal (sets=sets@entry=0x7ffe9cfd7140, nsets=<optimized out>, p=p@entry=0x55c12e34f430, result=result@entry=0x7ffe9cfd71a4) at /usr/src/debug/fontconfig-2.15.0-8.fc41.x86_64/src/fcmatch.c:964
#11 0x00007f2d68731b97 in IA__FcFontMatch (config=0x7f2d60000b70, config@entry=0x0, p=p@entry=0x55c12e34f430, result=result@entry=0x7ffe9cfd71a4) at /usr/src/debug/fontconfig-2.15.0-8.fc41.x86_64/src/fcmatch.c:1084
#12 0x00007f2d68c2c8fd in _cairo_ft_resolve_pattern (pattern=0x55c12e34f430, font_matrix=<optimized out>, ctm=<optimized out>, font_options=0x7ffe9cfd7560) at ../src/cairo-ft-font.c:4420
#13 _cairo_ft_font_face_get_implementation (abstract_face=0x55c12e34b090, font_matrix=<optimized out>, ctm=<optimized out>, options=0x7ffe9cfd7560) at ../src/cairo-ft-font.c:4087
#14 0x00007f2d68be8297 in cairo_scaled_font_create (font_face=0x55c12e3591b0, font_matrix=0x55c12e3a2710, ctm=0x7ffe9cfd7530, options=0x7ffe9cfd7560) at ../src/cairo-scaled-font.c:1174
#15 0x00007f2d68bb7b08 in _cairo_gstate_ensure_scaled_font (gstate=0x55c12e3a2690) at ../src/cairo-gstate.c:1918
#16 _cairo_gstate_ensure_scaled_font (gstate=gstate@entry=0x55c12e3a2690) at ../src/cairo-gstate.c:1897
#17 0x00007f2d68bb7bd9 in _cairo_gstate_get_scaled_font (gstate=0x55c12e3a2690, scaled_font=<synthetic pointer>) at ../src/cairo-gstate.c:1790
#18 _cairo_default_context_get_scaled_font (abstract_cr=<optimized out>) at ../src/cairo-default-context.c:1325
#19 0x00007f2d68c0d422 in cairo_text_extents (cr=0x55c12e3a2660, utf8=utf8@entry=0x55c113bd0066 "Verifying", extents=extents@entry=0x7ffe9cfd76e0) at ../src/cairo.c:3414
#20 0x000055c113bcbe72 in render_frame (surface=surface@entry=0x55c12e3f6c80) at ../render.c:169
#21 0x000055c113bcca6b in surface_frame_handle_done (callback=<optimized out>, time=<optimized out>, data=0x55c12e3f6c80) at ../main.c:189
#22 surface_frame_handle_done (data=0x55c12e3f6c80, callback=<optimized out>, time=<optimized out>) at ../main.c:176
#23 0x00007f2d68022056 in ffi_call_unix64 () at ../src/x86/unix64.S:104
#24 0x00007f2d6801d74d in ffi_call_int (cif=cif@entry=0x7ffe9cfd7940, fn=<optimized out>, rvalue=<optimized out>, avalue=<optimized out>, closure=closure@entry=0x0) at ../src/x86/ffi64.c:673
#25 0x00007f2d6802064e in ffi_call (cif=cif@entry=0x7ffe9cfd7940, fn=<optimized out>, rvalue=rvalue@entry=0x0, avalue=avalue@entry=0x7ffe9cfd7a10) at ../src/x86/ffi64.c:710
#26 0x00007f2d68aa910e in wl_closure_invoke (closure=closure@entry=0x55c12e40aa60, target=<optimized out>, target@entry=0x55c12e431460, opcode=opcode@entry=0, data=<optimized out>, flags=1) at ../src/connection.c:1228
#27 0x00007f2d68aa9979 in dispatch_event (display=display@entry=0x55c12e3025f0, queue=queue@entry=0x55c12e3026e8) at ../src/wayland-client.c:1670
#28 0x00007f2d68aa9d73 in dispatch_queue (display=0x55c12e3025f0, queue=0x55c12e3026e8) at ../src/wayland-client.c:1816
#29 wl_display_dispatch_queue_pending (display=0x55c12e3025f0, queue=0x55c12e3026e8) at ../src/wayland-client.c:2058
#30 0x00007f2d68aab964 in wl_display_dispatch_queue (display=<optimized out>, queue=<optimized out>) at ../src/wayland-client.c:2034
#31 0x00007f2d68aab980 in wl_display_dispatch (display=<optimized out>) at ../src/wayland-client.c:2101
#32 0x000055c113bca034 in display_in (fd=<optimized out>, mask=<optimized out>, data=<optimized out>) at ../main.c:1070
#33 0x000055c113bc6ca1 in loop_poll (loop=0x55c12e3bc590) at ../loop.c:100
#34 main (argc=<optimized out>, argv=<optimized out>) at ../main.c:1292

No other software using cairo/fontconfig is affected, just swaylock, which is why I'm reporting it here. fontconfig#318 is likely related.

[kennylevinsen]

swaylock doesn't directly interact with fontconfig - that's all handled inside cairo, and the lock for the cache is entirely internal to fontconfig, released by fontconfig's own cache functions on return.

The possible ways this could happen:

1. A prior call to fontconfig went through a path where fontconfig failed to unlock the mutex, deadlocking all future calls
2. A thread has been spawned for some reason, and made a call to fontconfig that ended up in an infinite loop under lock. Check with info threads.
3. A memory corruption has occurred and the mutex was not, in fact, locked.

I see at least one path where fontconfig fails to unlock the cache lock: If allocation fails in FcCacheInsert. But that wouldn't normally happen.

[kennylevinsen]

While I have no idea if that's your issue, MR for the missing cache unlock: https://gitlab.freedesktop.org/fontconfig/fontconfig/-/merge_requests/341/diffs