[PATCH] ARM: cacheflush: disallow pending signals during cacheflush

[Peter Maydell]

On 13 November 2014 11:26, Will Deacon <will.deacon at arm.com> wrote:
> Whilst I don't think this is the correct solution, I agree that there's
> a potential issue here. We could change the restart return value to
> -ERESTARTNOINTR instead, but I can imagine something like a periodic
> SIGALRM which could prevent a large cacheflush from ever completing.
> Do we actually care about making forward progress in such a scenario?
>
> It is interesting to note that this change has been in mainline since
> May last year without any reported issues. That could be down to a number
> of reasons:
>
>   (1) People are using old kernels on ARM
>
>   (2) Code doesn't check the return value from the cacheflush system call,
>       because it historically always returned 0

...and the documentation comment in the source code didn't say
anything about the syscall having a return value; it only
described the input parameters. I would actually be surprised
if any userspace caller of this syscall checked its return value
(the libgcc cacheflush function used by gcc's clear_cache builtin
doesn't, to pick one popularly used example).

>   (3) People are getting lucky with timing, as this is likely difficult
>       to hit

    (4) The resulting misbehaviour ("my JIT crashes occasionally and
        non-reproducibly at some point possibly some while after the
        cacheflush call") will be extremely hard to track back
        to this kernel change

> This leaves me with the following questions:
>
>   - Has this change been shown to break anything in practice?
>   - Can we change the internal return value to -ERESTARTNOINTR?
>   - What do we do about kernels that *do* return -EINTR? (>=3.12?)

My suggestion would be "treat this as a bugfix, put it into
stable kernels in the usual way (and assume distros will pick
it up if appropriate)".

>   - Can we get a manpage put together to describe this mess?

That would be nice :-)

-- PMM