Date: Mon, 05 Apr 2021 03:36:07 +0000 From: bugzilla-noreply@freebsd.org To: bugs@FreeBSD.org Subject: [Bug 254774] [rtld] dl_iterate_phdr: dlpi_tls_data should be the iterated module's TLS image instead of TLS initialization image Message-ID: <bug-254774-227-nkrdg7sBeG@https.bugs.freebsd.org/bugzilla/> In-Reply-To: <bug-254774-227@https.bugs.freebsd.org/bugzilla/> References: <bug-254774-227@https.bugs.freebsd.org/bugzilla/>
next in thread | previous in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D254774 --- Comment #4 from maskray <emacsray@gmail.com> --- I want to use dlpi_tls_data for static TLS blocks (from the main executable= and initially loaded shared objects). Working dynamic TLS blocks will be nice to have for consistency. I will try calling __tls_get_addr({tls_modid,0}) as a workaround, but it probably forces allocation (at least in glibc; I guess FreeBSD rtld/libc ma= y be similar) so slightly inferior (but not matters for most not-too-size-concer= ned applications). If you are interested, there is a complex story behind "why does sanitizer runtime need to know static/dynamic TLS blocks". In the end I think we may need libc APIs. Once I know the context sufficien= tly well, I shall start a discussion on https://www.openwall.com/lists/libc-coo= rd Currently I have written down some notes in https://maskray.me/blog/2021-02-14-all-about-thread-local-storage#why-does-= compiler-rt-need-to-know-tls-blocks Precise static TLS blocks tracking is more important than precise dynamic T= LS blocks. --=20 You are receiving this mail because: You are the assignee for the bug.=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-254774-227-nkrdg7sBeG>