Re: [PATCH v3 05/49] semihosting: Add target_strlen for softmmu-uaccess.h

[Peter Maydell]

On Sat, 21 May 2022 at 05:51, Richard Henderson
<richard.henderson@linaro.org> wrote:
>
> On 5/20/22 17:03, Richard Henderson wrote:
> > Mirror the interface of the user-only function of the same name.
> > Use probe_access_flags for the common case of ram, and
> > cpu_memory_rw_debug for the uncommon case of mmio.
> >
> > Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> > ---
> > v3: Use probe_access_flags (pmm)
>
> Ho hum.  MIPS still selects semihosting without TCG, which means that
> probe_access_flags is not present.
>
> At present I'm assuming that semihosting can't work with KVM, and that we 
> should disable
> the feature.
>
>
> r~
>
> > ---
> >   include/semihosting/softmmu-uaccess.h |  3 ++
> >   semihosting/uaccess.c                 | 49 +++++++++++++++++++++++++++
> >   2 files changed, 52 insertions(+)
> >
> > diff --git a/include/semihosting/softmmu-uaccess.h 
> > b/include/semihosting/softmmu-uaccess.h
> > index 03300376d3..4f08dfc098 100644
> > --- a/include/semihosting/softmmu-uaccess.h
> > +++ b/include/semihosting/softmmu-uaccess.h
> > @@ -53,4 +53,7 @@ void softmmu_unlock_user(CPUArchState *env, void *p,
> >                            target_ulong addr, target_ulong len);
> >   #define unlock_user(s, args, len) softmmu_unlock_user(env, s, args, len)
> >
> > +ssize_t softmmu_strlen_user(CPUArchState *env, target_ulong addr);
> > +#define target_strlen(p) softmmu_strlen_user(env, p)
> > +
> >   #endif /* SEMIHOSTING_SOFTMMU_UACCESS_H */
> > diff --git a/semihosting/uaccess.c b/semihosting/uaccess.c
> > index 0d3b32b75d..51019b79ff 100644
> > --- a/semihosting/uaccess.c
> > +++ b/semihosting/uaccess.c
> > @@ -8,6 +8,7 @@
> >    */
> >
> >   #include "qemu/osdep.h"
> > +#include "exec/exec-all.h"
> >   #include "semihosting/softmmu-uaccess.h"
> >
> >   void *softmmu_lock_user(CPUArchState *env, target_ulong addr,
> > @@ -23,6 +24,54 @@ void *softmmu_lock_user(CPUArchState *env, target_ulong 
> > addr,
> >       return p;
> >   }
> >
> > +ssize_t softmmu_strlen_user(CPUArchState *env, target_ulong addr)
> > +{
> > +    int mmu_idx = cpu_mmu_index(env, false);
> > +    size_t len = 0;
> > +
> > +    while (1) {
> > +        size_t left_in_page;
> > +        int flags;
> > +        void *h;
> > +
> > +        /* Find the number of bytes remaining in the page. */
> > +        left_in_page = -(addr | TARGET_PAGE_MASK);

This works but I'm never really a fan of expressions that mix
arithmetic and logical operations -- I always have to start writing
out bit patterns and checking what 2s-complement representation
involves in order to find out what they really do.

left_in_page = TARGET_PAGE_SIZE - (addr & ~TARGET_PAGE_MASK);

seems to me like it would be clearer.

You're correct that MIPS shouldn't be selecting semihosting in
a KVM-only compile -- there's no code to handle getting from the
guest executing the sdbbp insn to QEMU userspace to the code in
QEMU to deal with a semihosting call.

So for this patch:
Reviewed-by: Peter Maydell <peter.maydell@linaro.org>

-- PMM