On a 32-bit system, sizeof(mem_block) is 8, _upb_arena_alignup(sizeof(mem_block)) is 16.
In https://github.com/protocolbuffers/protobuf/blob/master/ruby/ext/google/protobuf_c/upb.c#L2361, alloc block size is 16384 + 8 bytes,
In https://github.com/protocolbuffers/protobuf/blob/master/ruby/ext/google/protobuf_c/upb.c#L2352, a->head.ptr moves forward 16 bytes.
So in https://github.com/protocolbuffers/protobuf/blob/master/ruby/ext/google/protobuf_c/upb.h#L380, if() is always false, and keeps calling malloc until infinite recursion(O0)/run out of memory(O3)

(gdb) f 16
#16 0x089fb8d0 in upb_arena_allocblock (a=0xa6bf14c, size=16384) at ../../../../ext/google/protobuf_c/upb.c:2361
2361 ../../../../ext/google/protobuf_c/upb.c: No such file or directory.
(gdb) p sizeof(mem_block)
$1 = 8
(gdb) p _upb_arena_alignup(sizeof(mem_block))
$2 = 16
(gdb) p UPB_MAX(size, a->next_block_size)
$3 = 16384
(gdb) f 18
#18 0x089f4d45 in upb_arena_malloc (a=0xa6bf14c, size=16384) at ../../../../ext/google/protobuf_c/upb.h:385
385 ../../../../ext/google/protobuf_c/upb.h: No such file or directory.
(gdb) p h->end - h->ptr
$4 = 16376
(gdb) p size
$5 = 16384

reference: gRPC-Ruby internal issue number b/171421508.
Issue description:
Segfault observed occasionally on Windows, from a cloud logging agent called google-fluentd, which captures logs from different processes on a VM and uses gRPC-Ruby to publish those logs message to a logging service.

How to reproduce:
0.Start a windows VM, e.x. windows 2016 datacenter. RDP into the VM then install the logging agent. Detailed Instruction here: https://cloud.google.com/logging/docs/agent/installation#joint-install

1. stop Stackdriver Logging service from Services.
2. clean folder LoggingAgent\Main\pos\winevtlog.pos\worker0\ and folder C:\var\log\google-fluentd\buffers.
3. start Stackdriver Logging service from Services.
4. stop Stackdriver Logging service from Services, when some large( over 300kb) log file appears in C:\var\log\google-fluentd\buffers\ folder
5. run Main\bin\ruby.exe 'Main\bin\fluentd' -c 'fluent.conf' -x stackdriverlogging -vvv
   Then Worker 0 finished with signal SIGSEGV. For any large( over 300kb) log file, SIGSEGV shows by 20% chance.
6. Once SIGSEGV has shown, as long as the log file remains, SIGSEGV shows every time when LoggingAgent runs.