Just wanted to share a small update on the situation after updating QEMU to the new Alpine package patched with Natanael's patch. So far so good, moreover I switched a few other guests from SATA to VirtIO SCSI and after two days no issues. Unless I find any problem I will report back with an update in a week from now.
Thanks everyone for all you did to help find a solution to this issue. On mar, feb 26, 2019 at 2:30 PM, Paolo Bonzini <pbonz...@redhat.com> wrote: On 23/02/19 12:49, Natanael Copa wrote: I suspect this happens due to the Alpine toolchain will enable _FORTIFY_SOURCE=2 by default and the way this is implemented via fortify-headers: http://git.2f30.org/fortify-headers/file/include/string.h.html#l39 The call to __orig_memcpy is the culprit there, is there any reason not to do something like _FORTIFY_FN(memcpy) void *__memcpy_chk(void *__od, const void *__os, size_t __n) { size_t __bd = __builtin_object_size(__od, 0); size_t __bs = __builtin_object_size(__os, 0); char *__d = (char *)__od; const char *__s = (const char *)__os; /* trap if pointers are overlapping but not if dst == src. * gcc seems to like to generate code that relies on dst == src */ if ((__d < __s && __d + __n > __s) || (__s < __d && __s + __n > __d)) __builtin_trap(); if (__n > __bd || __n > __bs) __builtin_trap(); return memcpy(__od, __os, __n); } #define memcpy __memcpy_chk ? That is, getting rid of _FORTIFY_ORIG altogether. Paolo
