Re: [PATCH] selftests/rseq: Play nice with binaries statically linked against glibc 2.35+

[Mathieu Desnoyers <mathieu.desnoyers@efficios.com>]

On 7/21/23 18:33, Sean Christopherson wrote:
> To allow running rseq and KVM's rseq selftests as statically linked
> binaries, initialize the various "trampoline" pointers to point directly
> at the expect glibc symbols, and skip the dlysm() lookups if the rseq
> size is non-zero, i.e. the binary is statically linked *and* the libc
> registered its own rseq.
> 
> Define weak versions of the symbols so as not to break linking against
> libc versions that don't support rseq in any capacity.
> 
> The KVM selftests in particular are often statically linked so that they
> can be run on targets with very limited runtime environments, i.e. test
> machines.

Thanks!

Reviewed-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

> 
> Fixes: 233e667e1ae3 ("selftests/rseq: Uplift rseq selftests for compatibility with glibc-2.35")
> Cc: Aaron Lewis <aaronlewis@google.com>
> Cc: kvm@vger.kernel.org
> Cc: stable@vger.kernel.org
> Signed-off-by: Sean Christopherson <seanjc@google.com>
> ---
> 
> Note, this is very much the result of throwing noodles until something
> stuck, it seems like there's gotta be a less awful way to handle this :-(
> 
> I Cc'd stable@ because I know I'm not the only person that runs statically
> linked KVM selftests, and figuring all this out was quite painful.
> 
>   tools/testing/selftests/rseq/rseq.c | 28 ++++++++++++++++++++++------
>   1 file changed, 22 insertions(+), 6 deletions(-)
> 
> diff --git a/tools/testing/selftests/rseq/rseq.c b/tools/testing/selftests/rseq/rseq.c
> index 4e4aa006004c..a723da253244 100644
> --- a/tools/testing/selftests/rseq/rseq.c
> +++ b/tools/testing/selftests/rseq/rseq.c
> @@ -34,9 +34,17 @@
>   #include "../kselftest.h"
>   #include "rseq.h"
>   
> -static const ptrdiff_t *libc_rseq_offset_p;
> -static const unsigned int *libc_rseq_size_p;
> -static const unsigned int *libc_rseq_flags_p;
> +/*
> + * Define weak versions to play nice with binaries that are statically linked
> + * against a libc that doesn't support registering its own rseq.
> + */
> +__weak ptrdiff_t __rseq_offset;
> +__weak unsigned int __rseq_size;
> +__weak unsigned int __rseq_flags;
> +
> +static const ptrdiff_t *libc_rseq_offset_p = &__rseq_offset;
> +static const unsigned int *libc_rseq_size_p = &__rseq_size;
> +static const unsigned int *libc_rseq_flags_p = &__rseq_flags;
>   
>   /* Offset from the thread pointer to the rseq area. */
>   ptrdiff_t rseq_offset;
> @@ -155,9 +163,17 @@ unsigned int get_rseq_feature_size(void)
>   static __attribute__((constructor))
>   void rseq_init(void)
>   {
> -	libc_rseq_offset_p = dlsym(RTLD_NEXT, "__rseq_offset");
> -	libc_rseq_size_p = dlsym(RTLD_NEXT, "__rseq_size");
> -	libc_rseq_flags_p = dlsym(RTLD_NEXT, "__rseq_flags");
> +	/*
> +	 * If the libc's registered rseq size isn't already valid, it may be
> +	 * because the binary is dynamically linked and not necessarily due to
> +	 * libc not having registered a restartable sequence.  Try to find the
> +	 * symbols if that's the case.
> +	 */
> +	if (!*libc_rseq_size_p) {
> +		libc_rseq_offset_p = dlsym(RTLD_NEXT, "__rseq_offset");
> +		libc_rseq_size_p = dlsym(RTLD_NEXT, "__rseq_size");
> +		libc_rseq_flags_p = dlsym(RTLD_NEXT, "__rseq_flags");
> +	}
>   	if (libc_rseq_size_p && libc_rseq_offset_p && libc_rseq_flags_p &&
>   			*libc_rseq_size_p != 0) {
>   		/* rseq registration owned by glibc */
> 
> base-commit: 88bb466c9dec4f70d682cf38c685324e7b1b3d60

-- 
Mathieu Desnoyers
EfficiOS Inc.
https://www.efficios.com