* bug#44326: 27.1.50; busy loop in lisp_free_align
2020-10-30 11:18 bug#44326: 27.1.50; busy loop in lisp_free_align Aaron Jensen
@ 2020-10-30 14:33 ` Aaron Jensen
2021-01-27 4:28 ` Lars Ingebrigtsen
0 siblings, 1 reply; 4+ messages in thread
From: Aaron Jensen @ 2020-10-30 14:33 UTC (permalink / raw)
To: 44326
Some more information that is hopefully helpful.
My Emacs memory usage is currently around 2.5GB.
Lisp Stack trace from busy loop:
(unsigned char *) $2 = 0x00000001003f48c6 “timer-event-handler”
(unsigned char *) $3 = 0x00000001003f074c “apply”
(unsigned char *) $4 = 0x00000001191ba690 “gcmh-idle-garbage-collect”
(unsigned char *) $5 = 0x00000001003ee2fe “garbage-collect”
GC Stats during busy loop. These don't seem to increase, it stays busy
in lisp_align_free, though with different blocks over time.
(lldb) p num_used
(object_ct) $8 = 2924
(lldb) p num_free
(object_ct) $9 = 13506
Here is a macOS sample from activity monitor:
Physical footprint: 2.5G
Physical footprint (peak): 3.2G
----
Call graph:
2102 Thread_6786366 DispatchQueue_1: com.apple.main-thread (serial)
+ 2102 start (in libdyld.dylib) + 1 [0x7fff6a33dcc9]
+ 2102 main (in emacs) + 6980 [0x100161764] emacs.c:2066
+ 2102 Frecursive_edit (in emacs) + 313 [0x100164299] keyboard.c:786
+ 2102 recursive_edit_1 (in emacs) + 192 [0x100163f50]
keyboard.c:714
+ 2102 command_loop (in emacs) + 202 [0x1001640ca]
keyboard.c:1070
+ 2102 internal_catch (in emacs) + 74 [0x1002614ba] eval.c:1117
+ 2102 command_loop_2 (in emacs) + 44 [0x10017ce8c]
keyboard.c:1091
+ 2102 internal_condition_case (in emacs) + 127
[0x100261b4f] eval.c:1356
+ 2102 command_loop_1 (in emacs) + 1449
[0x100165139] keyboard.c:1350
+ 2102 read_key_sequence (in emacs) + 1897
[0x100166719] keyboard.c:9554
+ 2102 read_char (in emacs) + 5298
[0x10016b5e2] keyboard.c:2738
+ 2102 sit_for (in emacs) + 750
[0x100008f8e] dispnew.c:6056
+ 2102 wait_reading_process_output (in
emacs) + 5631 [0x1002edbaf] process.c:5707
+ 2102 detect_input_pending_run_timers
(in emacs) + 54 [0x10016d056] keyboard.c:10368
+ 2102 get_input_pending (in emacs) +
64 [0x100170960] keyboard.c:6807
+ 2102 readable_events (in emacs) +
31 [0x10016e44f] keyboard.c:3397
+ 2102 timer_check (in emacs) +
168 [0x100170aa8] keyboard.c:4398
+ 2102 timer_check_2 (in emacs)
+ 1695 [0x1001711bf] keyboard.c:4336
+ 2102 call1 (in emacs) + 63
[0x100268d2f] eval.c:2655
+ 2102 Ffuncall (in emacs)
+ 542 [0x10026824e] eval.c:2797
+ 2102 funcall_lambda (in
emacs) + 508 [0x10026985c] eval.c:2990
+ 2102 exec_byte_code
(in emacs) + 8766 [0x1002d951e] bytecode.c:633
+ 2102 Ffuncall (in
emacs) + 468 [0x100268204] eval.c:2795
+ 2102 funcall_subr
(in emacs) + 267 [0x10026930b] eval.c:2848
+ 2102 Fapply (in
emacs) + 138 [0x1002649ca] eval.c:2378
+ 2102 Ffuncall
(in emacs) + 542 [0x10026824e] eval.c:2797
+ 2102
funcall_lambda (in emacs) + 508 [0x10026985c] eval.c:2990
+ 2102
exec_byte_code (in emacs) + 8766 [0x1002d951e] bytecode.c:633
+ 2102
Ffuncall (in emacs) + 468 [0x100268204] eval.c:2795
+ 2102
funcall_subr (in emacs) + 454 [0x1002693c6] eval.c:2866
+ 2102
Fgarbage_collect (in emacs) + 60 [0x10021f7ec] alloc.c:6056
+
2102 garbage_collect (in emacs) + 867 [0x10021e623] alloc.c:5984
+
2102 gc_sweep (in emacs) + 14 [0x10021f58e] alloc.c:7013
+
2090 sweep_conses (in emacs) + 652 [0x10022424c] alloc.c:6786
+
! 2034 lisp_align_free (in emacs) + 274,270,...
[0x10021c692,0x10021c68e,...] alloc.c:1245
+
! 32 lisp_align_free (in emacs) + 300,280,...
[0x10021c6ac,0x10021c698,...] alloc.c:1247
+
! 9 lisp_align_free (in emacs) + 376 [0x10021c6f8] alloc.c:1260
+
! : 7 free_small (in libsystem_malloc.dylib) + 1464
[0x7fff6a4f7d9e]
+
! : | 6 mvm_madvise_free (in libsystem_malloc.dylib) + 87
[0x7fff6a4fe93e]
+
! : | + 6 madvise (in libsystem_kernel.dylib) + 10 [0x7fff6a48104a]
+
! : | 1 mvm_madvise_free (in libsystem_malloc.dylib) + 8
[0x7fff6a4fe8ef]
+
! : 1 free (in libsystem_malloc.dylib) + 107 [0x7fff6a4f4a1c]
+
! : | 1 szone_size (in libsystem_malloc.dylib) + 73
[0x7fff6a4f73a1]
+
! : | 1 small_size (in libsystem_malloc.dylib) + 144
[0x7fff6a4f7636]
+
! : 1 free_small (in libsystem_malloc.dylib) + 2090
[0x7fff6a4f8010]
+
! : 1 small_free_scan_madvise_free (in libsystem_malloc.dylib) +
451 [0x7fff6a501e54]
+
! : 1 mvm_madvise_free (in libsystem_malloc.dylib) + 87
[0x7fff6a4fe93e]
+
! : 1 madvise (in libsystem_kernel.dylib) + 10
[0x7fff6a48104a]
+
! 7 lisp_align_free (in emacs) + 348,352 [0x10021c6dc,0x10021c6e0]
alloc.c:1253
+
! 5 lisp_align_free (in emacs) + 86 [0x10021c5d6] alloc.c:1228
+
! : 5 mem_find (in emacs) + 109 [0x10021ceed] alloc.c:3960
+
! 3 lisp_align_free (in emacs) + 94 [0x10021c5de] alloc.c:1228
+
! 3 mem_delete (in emacs) + 399 [0x10022200f] alloc.c:4220
+
! 2 xfree (in emacs) + 64 [0x100210a70] alloc.c:768
+
! | 1 free_tiny (in libsystem_malloc.dylib) + 459
[0x7fff6a4f8d8b]
+
! | + 1 tiny_free_no_lock (in libsystem_malloc.dylib) + 1111
[0x7fff6a4f92d6]
+
! | + 1 tiny_free_list_add_ptr (in libsystem_malloc.dylib) +
478 [0x7fff6a4f9ac8]
+
! | 1 free_tiny (in libsystem_malloc.dylib) + 479
[0x7fff6a4f8d9f]
+
! 1 xfree (in emacs) + 37 [0x100210a55] alloc.c:765
+
! 1 pdumper_object_p (in emacs) + 59 [0x100210abb]
pdumper.h:148
+
3 sweep_conses (in emacs) + 294,339,...
[0x1002240e6,0x100224113,...] alloc.c:6761
+
2 sweep_conses (in emacs) + 146 [0x100224052] alloc.c:6739
+
2 sweep_conses (in emacs) + 389 [0x100224145] alloc.c:6764
+
2 sweep_conses (in emacs) + 420 [0x100224164] alloc.c:6766
+
! 2 dead_object (in emacs) + 18 [0x10021aa32] lisp.h:1304
+
! 2 make_lisp_ptr (in emacs) + 0,8 [0x100219e90,0x100219e98]
lisp.h:1284
+
1 sweep_conses (in emacs) + 276 [0x1002240d4] alloc.c:6759
+
1 sweep_conses (in emacs) + 261 [0x1002240c5] alloc.c:6760
+
1 sweep_conses (in emacs) + 401 [0x100224151] alloc.c:6765
Aaron
On Fri, Oct 30, 2020 at 6:18 AM Aaron Jensen <aaronjensen@gmail.com> wrote:
>
> This should probably be merged into a reopened #30322
>
> I'm seeing a busy loop in lisp_free_align often on Emacs 27. It does
> eventually recover, but it takes a minute or more.
>
> * thread #1, queue = 'com.apple.main-thread', stop reason = signal SIGSTOP
> frame #0: 0x000000010021c692
> emacs`lisp_align_free(block=0x0000000291d2d400) at alloc.c:1245:7
> 1242 struct ablock **tem = &free_ablock;
> 1243 struct ablock *atop = &abase->blocks[aligned ?
> ABLOCKS_SIZE : ABLOCKS_SIZE - 1];
> 1244
> -> 1245 while (*tem)
> 1246 {
> 1247 if (*tem >= (struct ablock *) abase && *tem < atop)
> 1248 {
>
> I'm running gcmh-mode, and the last time it hung it was a gcmh-mode
> idle collection. Not sure if that makes any difference.
>
> Aaron
^ permalink raw reply [flat|nested] 4+ messages in thread