<div dir="ltr">
<p class="gmail-p1" style="margin:0px;font-variant-numeric:normal;font-variant-east-asian:normal;font-stretch:normal;font-size:11px;line-height:normal;font-family:Menlo;color:rgb(0,0,0)"><span style="color:rgb(34,34,34);font-family:Arial,Helvetica,sans-serif;font-size:small">Hi Mikael,</span></p><div><br></div><div>Please find flies you requested in attachment as <span style="font-variant-ligatures:no-common-ligatures;color:rgb(0,0,0);font-family:Menlo;font-size:11px">erl_files.tar.gz (compressed as facing issue with mail size)</span></div>
<div><br></div><div>Normal build option is:</div><div><br></div><div># gcc -Werror=undef -Werror=implicit -Werror=return-type -m64 -g -O3 -fomit-frame-pointer -Ierlang/src/solaris/otp/erts/sparc-sun-solaris2.10 -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -fno-tree-copyrename -DHAVE_CONFIG_H -Wall -Wstrict-prototypes -Wmissing-prototypes -Wdeclaration-after-statement -DUSE_THREADS -D_THREAD_SAFE -D_REENTRANT -DPOSIX_THREADS -D_POSIX_PTHREAD_SEMANTICS -Isparc-sun-solaris2.10/opt/smp -Ibeam -Isys/unix -Isys/common -Isparc-sun-solaris2.10 -Izlib -Ipcre -Ihipe -I../include -I../include/sparc-sun-solaris2.10 -I../include/internal -I../include/internal/sparc-sun-solaris2.10 -c beam/erl_alloc_util.c -o obj/sparc-sun-solaris2.10/opt/smp/erl_alloc_util.o<br><br>after your suggestion I updated it as below to generate erl_alloc_util file:</div><div><br># gcc -Werror=undef -Werror=implicit -Werror=return-type -m64 -g -O3 -fomit-frame-pointer -Ierlang/src/solaris/otp/erts/sparc-sun-solaris2.10 -D_LARGEFILE_SOURCE -D_FILE_OFFSET_BITS=64 -fno-tree-copyrename -DHAVE_CONFIG_H -Wall -Wstrict-prototypes -Wmissing-prototypes -Wdeclaration-after-statement -DUSE_THREADS -D_THREAD_SAFE -D_REENTRANT -DPOSIX_THREADS -D_POSIX_PTHREAD_SEMANTICS -Isparc-sun-solaris2.10/opt/smp -Ibeam -Isys/unix -Isys/common -Isparc-sun-solaris2.10 -Izlib -Ipcre -Ihipe -I../include -I../include/sparc-sun-solaris2.10 -I../include/internal -I../include/internal/sparc-sun-solaris2.10 -E beam/erl_alloc_util.c -o obj/sparc-sun-solaris2.10/opt/smp/erl_alloc_util.i<br></div><div><br></div><div>Also one thing I missed to mention, we are using <span style="font-family:"Segoe UI",system-ui,"Apple Color Emoji","Segoe UI Emoji",sans-serif;font-size:14px">gcc version 4.9.2 (GCC) for building on solaris SPARC as erlang doesn't support Sun's native compiler.</span></div><div><br></div><div>Thanks & Regards,</div><div>Pooja</div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Tue, May 12, 2020 at 10:44 PM Mikael Pettersson <<a href="mailto:mikpelinux@gmail.com">mikpelinux@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On Tue, May 12, 2020 at 4:18 PM Pooja Desai <<a href="mailto:pooja.desai10@gmail.com" target="_blank">pooja.desai10@gmail.com</a>> wrote:<br>
><br>
> Hi,<br>
><br>
><br>
><br>
> Thanks for response Mikael<br>
><br>
> As per your suggestion I am trying to write similar code to conclude if there is some issue with Solaris SPARC compiler.<br>
><br>
><br>
><br>
> But I have some doubts,<br>
><br>
> 1. If there is problem with compiler then we should be able to see this crash everywhere else also, any idea why its only reproduced here?<br>
><br>
> 2. As I understand your explanation it reads 64 bits by assembling two adjacent 32 bits fields. Will it really cause problem in multi-threaded program? Considering while context switching to another thread, OS will save current context of the thread (and hence registers) and will bring back when thread is active again.<br>
><br>
><br>
<br>
Breaking up a 64-bit load into two 32-bit loads loses atomicity with<br>
any concurrent store into that location, meaning the read may end up<br>
observing a result composed of 32 bit from the old value and 32 bit<br>
from the newly stored value, whereas the code expects to see either<br>
the old or the new, but never this mixture. This can happen also on a<br>
single-threaded CPU with preemptive multitasking.<br>
<br>
To move forward on the issue, I think you need to recreate the<br>
pre-processed source for erl_alloc_util.c. To do that:<br>
1. Compile Erlang/OTP as usual, starting from a pristine source<br>
directory (no left-overs from a previous build, best is to start fresh<br>
somewhere), but pass "V=1" to make. Save the output from "make" in a<br>
file.<br>
2. Note the step where it compiles erl_alloc_util.c.<br>
3. Reexecute that step, but replace any "-c" with "-E" and "-o<br>
erl_alloc_util.o" with "-o erl_alloc_util.i".<br>
4. Please send this ".i" file, together with the exact build steps and<br>
configuration options you used, and<br>
"erts/sparc-sun-solaris11/config.h" (I'm guessing the file name here)<br>
to me.<br>
<br>
My theory is that Erlang/OTP selects the wrong low-level primitives<br>
for this platform.<br>
<br>
<br>
><br>
><br>
> Thanks & Regards,<br>
><br>
> Pooja<br>
><br>
><br>
> On Mon, May 11, 2020 at 10:36 PM Mikael Pettersson <<a href="mailto:mikpelinux@gmail.com" target="_blank">mikpelinux@gmail.com</a>> wrote:<br>
>><br>
>> Hello Pooja,<br>
>><br>
>> On Mon, May 11, 2020 at 8:10 AM Pooja Desai <<a href="mailto:pooja.desai10@gmail.com" target="_blank">pooja.desai10@gmail.com</a>> wrote:<br>
>> ><br>
>> > Hi,<br>
>> ><br>
>> > Facing erlang core issue on solaris SPARC setup while running RabbitMQ<br>
>><br>
>> This looks like a 64-bit build, but the code doesn't look similar to<br>
>> what I get with gcc-9.3, so I'm assuming you used Sun's compiler?<br>
>><br>
>><br>
>> > (dbx) where<br>
>> ><br>
>> > =>[1] cpool_insert(0x1004efd40, 0xffffffff75600000, 0x61850, 0xffffffff75600018, 0x90f, 0x1004effd0), at 0x10006db14<br>
>> ><br>
>> > [2] abandon_carrier(0x1004efd40, 0xffffffff75600000, 0xffffffff75645ec0, 0xffffffff77d03818, 0x0, 0x6), at 0x10006de3c<br>
>> ><br>
>> > [3] 17(0x1004efd40, 0xcb3, 0x2, 0xffffffff75645e60, 0x0, 0x1004efd40), at 0x10006e958<br>
>> ><br>
>> > [4] erts_alcu_check_delayed_dealloc(0x1004efd40, 0x1, 0xffffffff77d03a40, 0xffffffff77d03a48, 0xffffffff77d03a44, 0x8000000000000007), at 0x100075244<br>
>> ><br>
>> > [5] erts_alloc_scheduler_handle_delayed_dealloc(0xffffffff3a82a620, 0xffffffff77d03a40, 0xffffffff77d03a48, 0xffffffff77d03a44, 0x100464, 0xffffffff3a82a5d0),<br>
>> ><br>
>> > at 0x1000622c0<br>
>> ><br>
>> > [6] handle_aux_work(0xffffffff3a8204a0, 0x2, 0x1, 0x2, 0x100400, 0x4e5ce123), at 0x1002a6044<br>
>> ><br>
>> > [7] erts_schedule(0xffffffff3a820380, 0x9, 0x9, 0xffffffff3a81fc80, 0x2, 0x2), at 0x1002a3040<br>
>> ><br>
>> > [8] process_main(0x100469, 0xffffffff3a302240, 0xfa0, 0x802a, 0xffffffff38f00438, 0x3), at 0x1002901bc<br>
>> ><br>
>> > [9] sched_thread_func(0xffffffff3a820380, 0x0, 0x0, 0xffffffff7a911240, 0x100000, 0x1), at 0x100038f08<br>
>> ><br>
>> > [10] thr_wrapper(0xffffffff7fffc278, 0x0, 0x0, 0x100289d48, 0xffffffff3a820380, 0x100038da0), at 0x100289dc8<br>
>> ><br>
>> ><br>
>> ><br>
>> > This issue is extremely intermittent so I am not able to reproduce it with debug build. But on our test setup I have seen this core twice only for solaris Sparc server for other servers (RHEL, Suse linux, Solarisx86, Windows etc.) with similar test environment things are working fine.<br>
>> ><br>
>> > In two instances when I faced this issue we are restarting Rabbitmq server. i.e. stop RabbitMQ and epmd then run startup script for rabbitmq. This performs 2 operations,<br>
>> ><br>
>> > First ping rabbitmq using "rabbitmqctl ping" to confirm rabbitmq is not already running ( I guess in background this will also start epmd) and then start rabbitmq-server in detached mode.<br>
>> ><br>
>> > Core is generated while starting this demon.<br>
>> ><br>
>> ><br>
>> > I checked code around abandon_carrier("<a href="https://github.com/erlang/otp/blame/master/erts/emulator/beam/erl_alloc_util.c" rel="noreferrer" target="_blank">https://github.com/erlang/otp/blame/master/erts/emulator/beam/erl_alloc_util.c</a>") but nothing changed in that area recently. So I am really clueless situation.<br>
>> ><br>
>> > Please le me know if anyone faced similar issue in past or have any idea around this. Using OTP version 22.2 and RabbitMQ version 3.7.23.<br>
>> ><br>
>> > Let me know any further information is required, pasting full core dump information below:<br>
>> ><br>
>> > debugging core file of beam.smp (64-bit) from hostname01<br>
>> > file: temp_dir/erlang/erts-10.6/bin/beam.smp<br>
>> > initial argv:<br>
>> > /temp_dir/erlang/erts-10.6/bin/beam.smp -- -root /temp_dir/<br>
>> > threading model: native threads<br>
>> > status: process terminated by SIGSEGV (Segmentation Fault), addr=<br>
>> > ffffffff004631b0<br>
>><br>
>> Ok, this tells us the address was unmapped. (It's not an alignment<br>
>> fault, another common issue on SPARC.)<br>
>><br>
>><br>
>> ><br>
>> > C++ symbol demangling enabled<br>
>> ><br>
>> > # stack<br>
>> ><br>
>> > cpool_insert+0xd0(10051c500, ffffffff7a400000, ffffffff7a441de8, ffffffff7c903818, 0, 23)<br>
>> > dealloc_block.part.17+0x1c0(10051c500, cb3, 2, ffffffff7a441d88, 0, 10051c500)<br>
>> > erts_alcu_check_delayed_dealloc+0xe4(10051c500, 1, ffffffff7c903a40, ffffffff7c903a48, ffffffff7c903a44, 8000000000000007)<br>
>> > erts_alloc_scheduler_handle_delayed_dealloc+0x34(ffffffff3b729c20, ffffffff7c903a40, ffffffff7c903a48, ffffffff7c903a44, 100464, ffffffff3b729bd0)<br>
>> > handle_aux_work+0xa50(ffffffff3b71faa0, 402, 1, 402, 100400, 42da0c68)<br>
>> > erts_schedule+0x192c(ffffffff3b71f980, 9, 9, ffffffff3b71f280, 402, 2)<br>
>> > process_main+0xc4(100469, ffffffff3b202240, fa0, ffffffff3b71f980, 241, 100294204)<br>
>> > sched_thread_func+0x168(ffffffff3b71f980, 0, 0, ffffffff39401a40, 100000, 1)<br>
>> > thr_wrapper+0x80(ffffffff7fffb318, 0, 0, 100289d48, ffffffff3b71f980, 100038da0)<br>
>> > libc.so.1`_lwp_start(0, 0, 0, 0, 0, 0)<br>
>> ><br>
>> > #############################################################################<br>
>> ><br>
>> > # registers<br>
>> ><br>
>> > %g0 = 0x0000000000000000 %l0 = 0xffffffff7a4307a0<br>
>> > %g1 = 0xffffffff004631a1 %l1 = 0x0000000000000000<br>
>> > %g2 = 0x0000000000000000 %l2 = 0x0000000000000000<br>
>> > %g3 = 0x000000010051c798 %l3 = 0x0000000000000000<br>
>> > %g4 = 0xffffffff004631a0 %l4 = 0x0000000000000000<br>
>> > %g5 = 0x00000001004631a0 beam.smp`firstfit_carrier_pool %l5 = 0x0000000000000000<br>
>><br>
>> This is interesting. Notice how the low 32-bits 004631a0 show up in<br>
>> three variations:<br>
>> 1. 00000001004631a0 beam.smp`firstfit_carrier_pool (the address of the<br>
>> firstfit_carrier_pool global variable)<br>
>> 2. ffffffff004631a0 (the above, but with the high 32 bits replaced<br>
>> with all-bits-one)<br>
>> 3. ffffffff004631a1 (the above, but with a tag in the low bit)<br>
>><br>
>> > %g6 = 0x0000000000000000 %l6 = 0x0000000000000000<br>
>> > %g7 = 0xffffffff39401a40 %l7 = 0x0000000000000000<br>
>> > %o0 = 0x000000010051c500 %i0 = 0x000000010051c500<br>
>> > %o1 = 0xffffffff7a400000 %i1 = 0xffffffff7a400000<br>
>> > %o2 = 0x00000000000676c0 %i2 = 0xffffffff7a441de8<br>
>> > %o3 = 0xffffffff7a400018 %i3 = 0xffffffff7c903818<br>
>> > %o4 = 0x00000000000007b9 %i4 = 0x0000000000000000<br>
>> > %o5 = 0x000000010051c790 %i5 = 0x0000000000000023<br>
>> > %o6 = 0xffffffff7c902eb1 %i6 = 0xffffffff7c902f61<br>
>> > %o7 = 0x000000010006de3c abandon_carrier+0x118 %i7 = 0x000000010006e958 dealloc_block.part.17+0x1c0<br>
>> ><br>
>> > %ccr = 0x44 xcc=nZvc icc=nZvc<br>
>> > %y = 0x0000000000000000<br>
>> > %pc = 0x000000010006db14 cpool_insert+0xd0<br>
>> > %npc = 0x000000010006db18 cpool_insert+0xd4<br>
>> > %sp = 0xffffffff7c902eb1<br>
>> > %fp = 0xffffffff7c902f61<br>
>> ><br>
>> > %asi = 0x82<br>
>> > %fprs = 0x00<br>
>> ><br>
>> > # dissassembly around pc<br>
>> ><br>
>> > cpool_insert+0xa8: mov %g1, %g2<br>
>> > cpool_insert+0xac: ldx [%g5 + 0x10], %g1<br>
>> > cpool_insert+0xb0: membar #LoadLoad|#LoadStore<br>
>> > cpool_insert+0xb4: ba,pt %xcc, +0x1c <cpool_insert+0xd0><br>
>> > cpool_insert+0xb8: and %g1, -0x4, %g4<br>
>><br>
>> > cpool_insert+0xbc: membar #LoadLoad|#LoadStore<br>
>> > cpool_insert+0xc0: and %g2, 0x3, %g3<br>
>> > cpool_insert+0xc4: brz,pn %g3, +0x1ec <cpool_insert+0x2b0><br>
>> > cpool_insert+0xc8: mov %g2, %g1<br>
>> > cpool_insert+0xcc: and %g1, -0x4, %g4<br>
>> > cpool_insert+0xd0: ld [%g4 + 0x10], %g1<br>
>><br>
>> This is the faulting instruction. We're in the /* Find a predecessor<br>
>> to be, and set mod marker on its next ptr */ loop.<br>
>><br>
>> > cpool_insert+0xd4: ld [%g4 + 0x14], %g2<br>
>> > cpool_insert+0xd8: sllx %g1, 0x20, %g1<br>
>> > cpool_insert+0xdc: cmp %g5, %g4<br>
>> > cpool_insert+0xe0: bne,pt %xcc, -0x24 <cpool_insert+0xbc><br>
>> > cpool_insert+0xe4: or %g2, %g1, %g2<br>
>><br>
>> The above reads a 64-bit "->next" pointer by assembling two adjacent<br>
>> 32-bit fields. Weird, but arithmetically Ok.<br>
>><br>
>> Two things strike me:<br>
>> 1. The compiler implements "atomic load of 64-bits" as "load 32 bits,<br>
>> load another 32 bits, combine", which isn't correct in a multithreaded<br>
>> program. The error could be in the compiler, or in the source code.<br>
>> 2. In the register dump it was obvious that the high bits of an<br>
>> address had been clobbered.<br>
>><br>
>> My suspicion is that either Sun's compiler is buggy, or Erlang is<br>
>> selecting non thread-safe code in this case.<br>
>><br>
>> On SPARC64 Linux w/ GCC I get very different code that uses "ldx" for<br>
>> those 64-bit loads, as expected.<br>
>><br>
>> /Mikael<br>
</blockquote></div>