Re: [MTASCsft PATCH 26/??] MT-, AS- and AC-Safety docs: manual/setjmp.texi

["Carlos O'Donell" <carlos at redhat dot com>]

On 01/27/2014 10:03 PM, Alexandre Oliva wrote:
> The *context family of functions does not interact very well with the
> cleanup stack and thus with longjmp, but although this may break
> threads, it's not a *multi* thread safety issue; it could happen even
> with a single thread.
> 
> 
> for ChangeLog
> 
> 	* manual/setjmp.texi: Document MTASC-safety properties.

OK to checkin.

We can revisit your setcontext comments at a later time incrementally
when we do another round of edits to the safety notes for POSIX
conformance. Thanks for the notes, very invaluable.

> ---
>  manual/setjmp.texi |   44 ++++++++++++++++++++++++++++++++++++++++++++
>  1 file changed, 44 insertions(+)
> 
> diff --git a/manual/setjmp.texi b/manual/setjmp.texi
> index b3c0a7b..b924d58 100644
> --- a/manual/setjmp.texi
> +++ b/manual/setjmp.texi
> @@ -107,6 +107,10 @@ identify a specific place to return to.
>  @comment setjmp.h
>  @comment ISO
>  @deftypefn Macro int setjmp (jmp_buf @var{state})
> +@safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}}
> +@c _setjmp ok
> +@c  __sigsetjmp(!savemask) ok
> +@c   __sigjmp_save(!savemask) ok, does not call sigprocmask
>  When called normally, @code{setjmp} stores information about the
>  execution state of the program in @var{state} and returns zero.  If
>  @code{longjmp} is later used to perform a non-local exit to this
> @@ -116,6 +120,20 @@ execution state of the program in @var{state} and returns zero.  If
>  @comment setjmp.h
>  @comment ISO
>  @deftypefun void longjmp (jmp_buf @var{state}, int @var{value})
> +@safety{@prelim{}@mtsafe{}@asunsafe{@ascuplugin{} @asucorrupt{} @asulock{/hurd}}@acunsafe{@acucorrupt{} @aculock{/hurd}}}
> +@c __libc_siglongjmp @ascuplugin @asucorrupt @asulock/hurd @acucorrupt @aculock/hurd
> +@c  _longjmp_unwind @ascuplugin @asucorrupt @acucorrupt
> +@c   __pthread_cleanup_upto @ascuplugin @asucorrupt @acucorrupt
> +@c     plugins may be unsafe themselves, but even if they weren't, this
> +@c     function isn't robust WRT async signals and cancellation:
> +@c     cleanups aren't taken off the stack right away, only after all
> +@c     cleanups have been run.  This means that async-cancelling
> +@c     longjmp, or interrupting longjmp with an async signal handler
> +@c     that calls longjmp may run the same cleanups multiple times.
> +@c    _JMPBUF_UNWINDS_ADJ ok
> +@c    *cleanup_buf->__routine @ascuplugin
> +@c  sigprocmask(SIG_SETMASK) dup @asulock/hurd @aculock/hurd
> +@c  __longjmp ok
>  This function restores current execution to the state saved in
>  @var{state}, and continues execution from the call to @code{setjmp} that
>  established that return point.  Returning from @code{setjmp} by means of
> @@ -199,6 +217,11 @@ information about the set of blocked signals.
>  @comment setjmp.h
>  @comment POSIX.1
>  @deftypefun int sigsetjmp (sigjmp_buf @var{state}, int @var{savesigs})
> +@safety{@prelim{}@mtsafe{}@asunsafe{@asulock{/hurd}}@acunsafe{@aculock{/hurd}}}
> +@c sigsetjmp @asulock/hurd @aculock/hurd
> +@c  __sigsetjmp(savemask) @asulock/hurd @aculock/hurd
> +@c   __sigjmp_save(savemask) @asulock/hurd @aculock/hurd
> +@c    sigprocmask(SIG_BLOCK probe) dup @asulock/hurd @aculock/hurd
>  This is similar to @code{setjmp}.  If @var{savesigs} is nonzero, the set
>  of blocked signals is saved in @var{state} and will be restored if a
>  @code{siglongjmp} is later performed with this @var{state}.
> @@ -207,6 +230,8 @@ of blocked signals is saved in @var{state} and will be restored if a
>  @comment setjmp.h
>  @comment POSIX.1
>  @deftypefun void siglongjmp (sigjmp_buf @var{state}, int @var{value})
> +@safety{@prelim{}@mtsafe{}@asunsafe{@ascuplugin{} @asucorrupt{} @asulock{/hurd}}@acunsafe{@acucorrupt{} @aculock{/hurd}}}
> +@c Alias to longjmp.
>  This is similar to @code{longjmp} except for the type of its @var{state}
>  argument.  If the @code{sigsetjmp} call that set this @var{state} used a
>  nonzero @var{savesigs} flag, @code{siglongjmp} also restores the set of
> @@ -267,6 +292,10 @@ and modification happens through one of the following functions:
>  @comment ucontext.h
>  @comment SVID
>  @deftypefun int getcontext (ucontext_t *@var{ucp})
> +@safety{@prelim{}@mtsafe{@mtsrace{:ucp}}@assafe{}@acsafe{}}
> +@c Linux-only implementations in assembly, including sigprocmask
> +@c syscall.  A few cases call the sigprocmask function, but that's safe
> +@c too.  The ppc case is implemented in terms of a swapcontext syscall.
>  The @code{getcontext} function initializes the variable pointed to by
>  @var{ucp} with the context of the calling thread.  The context contains
>  the content of the registers, the signal mask, and the current stack.
> @@ -293,6 +322,8 @@ used to do that.
>  @comment ucontext.h
>  @comment SVID
>  @deftypefun void makecontext (ucontext_t *@var{ucp}, void (*@var{func}) (void), int @var{argc}, @dots{})
> +@safety{@prelim{}@mtsafe{@mtsrace{:ucp}}@assafe{}@acsafe{}}
> +@c Linux-only implementations mostly in assembly, nothing unsafe.
>  
>  The @var{ucp} parameter passed to the @code{makecontext} shall be
>  initialized by a call to @code{getcontext}.  The context will be
> @@ -339,6 +370,15 @@ requires detection of the platform at compile time.
>  @comment ucontext.h
>  @comment SVID
>  @deftypefun int setcontext (const ucontext_t *@var{ucp})
> +@safety{@prelim{}@mtsafe{@mtsrace{:ucp}}@asunsafe{@asucorrupt{}}@acunsafe{@acucorrupt{}}}
> +@c Linux-only implementations mostly in assembly.  Some ports use
> +@c sigreturn or swapcontext syscalls; others restore the signal mask
> +@c first and then proceed restore other registers in userland, which
> +@c leaves a window for cancellation or async signals with misaligned or
> +@c otherwise corrupt stack.  ??? Switching to a different stack, or even
> +@c to an earlier state on the same stack, may conflict with pthread
> +@c cleanups.  This is not quite MT-Unsafe, it's a different kind of
> +@c safety issue.
>  
>  The @code{setcontext} function restores the context described by
>  @var{ucp}.  The context is not modified and can be reused as often as
> @@ -372,6 +412,10 @@ there are situations where the current context has to be preserved.
>  @comment ucontext.h
>  @comment SVID
>  @deftypefun int swapcontext (ucontext_t *restrict @var{oucp}, const ucontext_t *restrict @var{ucp})
> +@safety{@prelim{}@mtsafe{@mtsrace{:oucp} @mtsrace{:ucp}}@asunsafe{@asucorrupt{}}@acunsafe{@acucorrupt{}}}
> +@c Linux-only implementations mostly in assembly.  Some ports call or
> +@c inline getcontext and/or setcontext, adjusting the saved context in
> +@c between, so we inherit the potential issues of both.
>  
>  The @code{swapcontext} function is similar to @code{setcontext} but
>  instead of just replacing the current context the latter is first saved
> 
>