Re: [PATCH] aarch64: Remove ld.so __tls_get_addr plt usage

[Szabolcs Nagy <szabolcs.nagy@arm.com>]

The 04/08/2024 13:57, Adhemerval Zanella Netto wrote:
> On 08/04/24 04:26, Szabolcs Nagy wrote:
> > The 04/07/2024 16:29, Cristian Rodríguez wrote:
> >> On Fri, Apr 5, 2024 at 11:59 AM Szabolcs Nagy <szabolcs.nagy@arm.com> wrote:
> >>> The 04/05/2024 09:35, Adhemerval Zanella wrote:
> >>>> Use the hidden alias instead.
> >>>>
> >>>> Checked on aarch64-linux-gnu.
> >>>
> >>> does this change behaviour in case __tls_get_addr is interposed?
> >>
> >> Wut ? is that really supported.. I mean.. isn't that symbol prefix
> >> reserved for the implementation and any assumption about it is either
> >> ID or UB?
> > 
> > a behaviour can change even if it's not supported.
> > i did not try to imply that it should be supported.
> > 
> > i know sanitizers interpose __tls_get_addr, because
> > https://sourceware.org/bugzilla/show_bug.cgi?id=16291
> > i don't know if that hack works at all now for tlsdesc
> > (where the ld.so calls __tls_get_addr, not user code)
> > 
> > my question was if we investigated this issue since it
> > is useful to document then in the commit msg (or news
> > entry if this affects users)
> 
> This change 'breaks' the sanitizer trick to get the dynamic TLS, with
> this patch I now see:
> 
>   MemorySanitizer-AARCH64 :: dtls_test.c
>   SanitizerCommon-asan-aarch64-Linux :: Linux/resize_tls_dynamic.cpp
>   SanitizerCommon-msan-aarch64-Linux :: Linux/resize_tls_dynamic.cpp
>   SanitizerCommon-tsan-aarch64-Linux :: Linux/resize_tls_dynamic.cpp
> 
> And it does not fail on x86 only because it uses -mtls=gnu as default
> (the same tests fail on x86 with -mtls=gnu2).
> 
> Now that GCC and distributions are aiming to use GNU2/DESC as the
> default TLS, this hack will also break on x86.  So the question is 
> whether we revert 050f7298e1ecc39887c329037575ccd972071255 and 
> document that __tls_get_addr should be interposable, or move with this 
> change and try to come up with a possible solution for BZ#16291.
> 
> I bringing this because we will have another two ABIs with tlsdesc
> support (loongarch and riscv).

adding some sanitizer committers to cc.

tl;dr: in the next glibc release tlsdesc will not call
__tls_get_addr in an interposable way in the dynamic tls
allocation case, unless somebody screems that this is needed.
(affects targets that may default to tlsdesc, but note that
the dynamic case only triggers with tlsdesc when a lot of
dlopened tls is used, otherwise static tls area is used)

i think it is also possible that we will use custom malloc
in ld.so which may be just as big change for the sanitizers.
(this can make tls access signal safe)

i'm not against the change, but if we plan to add several
interposable hooks as in
https://sourceware.org/glibc/wiki/ThreadPropertiesAPI
then we might as well keep __tls_get_addr PLT for now.