From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Return-Path: From: Cyrill Gorcunov Subject: [PATCH v5 0/2] Implement support of strip_core functionality Date: Fri, 17 May 2019 00:45:57 +0300 Message-Id: <20190516214559.3310-1-gorcunov@gmail.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit To: tml Cc: Vladimir Davydov , Alexander Turenko , gorcunov@gmail.com List-ID: The series patches both "small" memory engine and tarantool code as well, moreover the tarantool patch depends on "small" new api. Please review carefully. To test if memory doesn't go into coredump I've been using gcore together with readelf utilities. When strip_core is set to false (by default) 2019-05-16 23:53:28.755 [29583] main/102/interactive D> tuple arena memtx: addr 0x7f728c000000 size 268435456 flags 0x80000001 dontdump 0 2019-05-16 23:53:28.755 [29583] main/102/interactive I> mapping 134217728 bytes for vinyl tuple arena... 7f7283800000-7f729c081000 rw-p 00000000 00:00 0 VmFlags: rd wr mr mw me ac sd [34] load PROGBITS 00007f7283800000 0df6e284 0000000018881000 0000000000000000 WA 0 0 1 The segment with memtx memory is present inside corefile (because vma flags are the same for both memtx and vynil os does merge the memory into single vma). In turn when strip_core is set to true we have 2019-05-16 23:58:26.796 [29852] main/102/interactive D> tuple arena memtx: addr 0x7f6978000000 size 268435456 flags 0x80000005 dontdump 1 2019-05-16 23:58:26.796 [29852] main/102/interactive I> mapping 134217728 bytes for vinyl tuple arena... 7f6978000000-7f6979000000 rw-p 00000000 00:00 0 VmFlags: rd wr mr mw me ac dd sd [36] load PROGBITS 00007f6979000000 15f7d364 000000000f000000 0000000000000000 WA 0 0 1 The memtx is not present in corefile but only vynil goes there. Volodya, could you please help with more intensive testing since I'm not sure how to fill memtx from console or some script yet. Cyrill Gorcunov (2): slab_arena: Enhance slab_arena_create to support madvise hints box/memtx: Allow to skip tuple memory from coredump -- 2.20.1