Re: [PULL 02/60] semihosting: Return failure from softmmu-uaccess.h functions

[Richard Henderson]

On 7/29/22 07:31, Peter Maydell wrote:
> On Tue, 28 Jun 2022 at 05:54, Richard Henderson
> <richard.henderson@linaro.org> wrote:
> > We were reporting unconditional success for these functions;
> > pass on any failure from cpu_memory_rw_debug.
> >
> > Reviewed-by: Peter Maydell <peter.maydell@linaro.org>
> > Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
>
> So, this commit makes us check the cpu_memory_rw_debug()
> return value in the softmmu_lock_user() function; but Coverity
> points out (CID 1490294) that we still aren't checking the
> return value from when we call it in softmmu_unlock_user()...
>
> What, if anything, should we do about that? In particular,
> I think that for semihosting calls that write to guest memory,
> if the guest passes us a pointer to read-only memory we're
> not going to find out about that until unlock time.

Not even then, because address_space_write_rom will in fact write to rom, nevermind 
virtual page permissions.  Moreover, there are no failure conditions from 
address_space_write_rom.  It skips non-{ram,rom} and always returns OK.

It's probable that we should be using cpu_memory_rw_debug at all, but should be using a 
higher-level interface, something that (1) checks the virtual page permissions and (2) 
probably rejects semihosting to mmio.

But in the short term, I think we can just ignore the warning.


r~