Re: lib/54017: jemalloc deadlock?

[Rin Okuyama <rokuyama.rk%gmail.com@localhost>]

The following reply was made to PR lib/54017; it has been noted by GNATS.

From: Rin Okuyama <rokuyama.rk%gmail.com@localhost>
To: "gnats-bugs%NetBSD.org@localhost" <gnats-bugs%NetBSD.org@localhost>
Cc: 
Subject: Re: lib/54017: jemalloc deadlock?
Date: Mon, 11 Mar 2019 14:38:46 +0900

 Christos, thank you so much for your hard work!
 
 Now, deadlock with "make configure" in pkgsrc/www/firefox
 disappears on aarch64 (at least for > 10 tries).
 
 For problems reported by wiz:
 
 (1) Build failure of Thunderbird due to multiple definition of
 `malloc_foo': already fixed.
 
 (2) Build failure of Firefox due to stuck in cargo:
 
 I guess this is not due to new jemalloc. I have observed similar
 failures with old jemalloc; cargo sometimes deadlock randomly,
 and stop it by ^C and re-run make fixes the problem. Also, no
 assertion failures are detected with new jemalloc with
 JEMALLOC_DEBUG.
 
 (3) Build failure of pkgsrc/wip/emacs-git:
 
 With JEMALLOC_DEBUG enabled, it crashes earlier:
 
 ----
 Loading window (source)...
 <jemalloc>: /build/src/external/bsd/jemalloc/lib/../dist/src/arena.c:271: Failed assertion: "bitmap_get(slab_data->bitmap, &bin_info->bitmap_info, regind)"
 gmake[1]: *** [Makefile:810: bootstrap-emacs.pdmp] Abort trap (core dumped)
 gmake[1]: Leaving directory '/build/pkgsrc/wip/emacs-git/work.x86_64/emacs/src'
 gmake: *** [Makefile:423: src] Error 2
 *** Error code 2
 ----
 
 However, it is unclear whether this is due to bug in new jemalloc
 or that in emacs. Note that pkgsrc/editors/emacs can be build and
 works fine with new jemalloc.
 
 For test failures:
 
 https://mail-index.netbsd.org/current-users/2019/03/10/msg035279.html
 
 I do not observe tmpfs ones on amd64, however test for libc causes
 kernel panic on amd64
 
 ----
 pmap_unwire: wiring for pmap 0xffffa08d42f93d28 va 0x7827c6b86000did not change!
 panic: kernel diagnostic assertion "pg->wire_count > 0" failed: file "../../../.
 ./uvm/uvm_page.c", line 1594
 cpu8: Begin traceback...
 vpanic() at netbsd:vpanic+0x143
 stge_eeprom_wait.isra.4() at netbsd:stge_eeprom_wait.isra.4
 uvm_fault_upper_enter.isra.5() at netbsd:uvm_fault_upper_enter.isra.5+0x1c6
 uvm_fault_internal() at netbsd:uvm_fault_internal+0x14ae
 uvm_fault_wire() at netbsd:uvm_fault_wire+0x56
 uvm_map_pageable() at netbsd:uvm_map_pageable+0x220
 sys_mlock() at netbsd:sys_mlock+0xa3
 syscall() at netbsd:syscall+0x15b
 --- syscall (number 203) ---
 7827c604307a:
 cpu8: End traceback...
 uvm_fault(0xffffa08cbc513008, 0x0, 2) -> e
 ----
 
 in a similar manner to i386
 
 https://releng.netbsd.org/b5reports/i386/2019/2019.03.10.19.47.03/test.log
 
 I don't know this is related to jemalloc or not.
 
 Thanks,
 rin