just some findings (after the hint about jemalloc) The breaking commit is probably " 6328. [func] Add workaround to enforce dynamic linker to pull jemalloc earlier than libc to ensure all memory allocations are done via jemalloc. [GL #4404] " checking with gdb on older bind9 malloc and free point to glibc's implementation, on the current version to jemalloc.
Now I don't know who is to blame and who is responsible for this regression (bind or samba), but it is really strange why in some random function in samba's module malloc is "redirected" to jemalloc, but just one free insists to call glibc (I've checked, it is definitly a call to a free, not an explicit glibc free of some kind). Maybe the same issue as here https://github.com/jemalloc/jemalloc/issues/2419[1] since ldb_modules.c uses RTLD_DEEPBIND If that's the case then is this even fixable? Samba's upstream bug is here: https://bugzilla.samba.org/show_bug.cgi?id=15643[2] -------- [1] https://github.com/jemalloc/jemalloc/issues/2419 [2] https://bugzilla.samba.org/show_bug.cgi?id=15643