Re: [RFC PATCH glibc 4/8] glibc: Perform rseq(2) registration at C startup and thread creation (v15)

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

----- On Mar 20, 2020, at 9:44 AM, Mathieu Desnoyers mathieu.desnoyers@efficios.com wrote:
[...]
> Actually, here is an important clarification: the Linux kernel validates
> the struct rseq alignment on registration:
> 
>        if (!IS_ALIGNED((unsigned long)rseq, __alignof__(*rseq)) ||
>            rseq_len != sizeof(*rseq))
>                return -EINVAL;
> 
> So removing the aligned attribute from struct rseq is actually an
> ABI-breaking change, because it would be incompatible with older
> kernels which perform the IS_ALIGNED check expecting at least at
> 32 bytes alignment.

So I plan to add the following to glibc's sys/rseq.h:

#include <sys/cdefs.h>

[...]

/* Ensure the compiler supports __attribute__ ((aligned)).  */
_Static_assert (__alignof__ (struct rseq_cs) >= 4 * sizeof(uint64_t),
                "alignment");
_Static_assert (__alignof__ (struct rseq) >= 4 * sizeof(uint64_t),
                "alignment");

/* Allocations of struct rseq and struct rseq_cs on the heap need to
   be aligned on 32 bytes. Therefore, use of malloc is discouraged
   because it does not guarantee alignment. posix_memalign should be
   used instead.  */

Does it help mitigating your concerns ?

Thanks,

Mathieu

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