Line 698 of fs/binfmt_elf.c appears to contain a bug, whereby if a large non-BSS
(nobits) PT_LOAD segment is loaded set_brk is called which in turn calls do_brk,
which maps the area but discards the protection flags requested in the binary.
This bug was reported to me second-hand, so I can't be sure I got all the
details correct. Please let me know if not. This bug prevents Wine from using a
new ELF section to reserve parts of the address space that are required.
Any update on this problem? How are new kernels work for you?
Someone affiliated with Wine said that this problem has been worked around so far.
Do we need a proper fix to it?
To get a useful response, the bug needs details such as the eu-readelf -l layout of a binary that reproduces the problem.