This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] hppa: Fix stack alignment passed to clone


On 2019-10-19 4:07 p.m., Florian Weimer wrote:
> I think an unaligned stack triggers undefined behavior, so a crash
> would be acceptable in this context.  We could maybe add an assert to
> the thread creation (although that will cause backwards compatibility
> concerns on i386 due to the silent ABI change for SSE2).
The pthread_create() function is supposed to return EINVAL if the attributes specified by
the attr argument are invalid.  The Open Group specification doesn't say a crash is acceptable
behavior.  We don't actually have a "stack" at this point.  We have a pointer to memory to
be used as the stack for the new thread.

As I see it, we have two alternatives in the hppa __clone() implementation:

1) Return EINVAL when passed a misaligned pointer, or
2) Align the pointer.

Dave

-- 
John David Anglin  dave.anglin@bell.net



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]