[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [PATCH v3 4/5] common/domain: add a domain context record for shared_info...



> -----Original Message-----
> From: Jan Beulich <jbeulich@xxxxxxxx>
> Sent: 19 May 2020 15:08
> To: Paul Durrant <paul@xxxxxxx>
> Cc: xen-devel@xxxxxxxxxxxxxxxxxxxx; Paul Durrant <pdurrant@xxxxxxxxxx>; Ian 
> Jackson
> <ian.jackson@xxxxxxxxxxxxx>; Wei Liu <wl@xxxxxxx>; Andrew Cooper 
> <andrew.cooper3@xxxxxxxxxx>; George
> Dunlap <george.dunlap@xxxxxxxxxx>; Julien Grall <julien@xxxxxxx>; Stefano 
> Stabellini
> <sstabellini@xxxxxxxxxx>
> Subject: Re: [PATCH v3 4/5] common/domain: add a domain context record for 
> shared_info...
> 
> On 14.05.2020 12:44, Paul Durrant wrote:
> > @@ -61,6 +62,76 @@ static void dump_header(void)
> >
> >  }
> >
> > +static void print_binary(const char *prefix, void *val, size_t size,
> 
> const also for val?

Yes, it can be.

> 
> > +                         const char *suffix)
> > +{
> > +    printf("%s", prefix);
> > +
> > +    while (size--)
> 
> Judging from style elsewhere you look to be missing two blanks
> here.
> 

Yes.

> > +    {
> > +        uint8_t octet = *(uint8_t *)val++;
> 
> Following the above then also better don't cast const away here.
> 
> > +        unsigned int i;
> > +
> > +        for ( i = 0; i < 8; i++ )
> > +        {
> > +            printf("%u", octet & 1);
> > +            octet >>= 1;
> > +        }
> > +    }
> > +
> > +    printf("%s", suffix);
> > +}
> > +
> > +static void dump_shared_info(void)
> > +{
> > +    DOMAIN_SAVE_TYPE(SHARED_INFO) *s;
> > +    shared_info_any_t *info;
> > +    unsigned int i;
> > +
> > +    GET_PTR(s);
> > +
> > +    printf("    SHARED_INFO: has_32bit_shinfo: %s buffer_size: %u\n",
> > +           s->has_32bit_shinfo ? "true" : "false", s->buffer_size);
> > +
> > +    info = (shared_info_any_t *)s->buffer;
> > +
> > +#define GET_FIELD_PTR(_f) \
> > +    (s->has_32bit_shinfo ? (void *)&(info->x32._f) : (void 
> > *)&(info->x64._f))
> 
> Better cast to const void * ?
> 

Ok.

> > +#define GET_FIELD_SIZE(_f) \
> > +    (s->has_32bit_shinfo ? sizeof(info->x32._f) : sizeof(info->x64._f))
> > +#define GET_FIELD(_f) \
> > +    (s->has_32bit_shinfo ? info->x32._f : info->x64._f)
> > +
> > +    /* Array lengths are the same for 32-bit and 64-bit shared info */
> 
> Not really, no:
> 
>     xen_ulong_t evtchn_pending[sizeof(xen_ulong_t) * 8];
>     xen_ulong_t evtchn_mask[sizeof(xen_ulong_t) * 8];
> 

Oh, I must have misread.

> > @@ -167,12 +238,14 @@ int main(int argc, char **argv)
> >          if ( (typecode < 0 || typecode == desc->typecode) &&
> >               (instance < 0 || instance == desc->instance) )
> >          {
> > +
> >              printf("[%u] type: %u instance: %u length: %u\n", entry++,
> >                     desc->typecode, desc->instance, desc->length);
> 
> Stray insertion of a blank line?
> 

Yes.

> > @@ -1649,6 +1650,65 @@ int continue_hypercall_on_cpu(
> >      return 0;
> >  }
> >
> > +static int save_shared_info(const struct domain *d, struct domain_context 
> > *c,
> > +                            bool dry_run)
> > +{
> > +    struct domain_shared_info_context ctxt = { .buffer_size = PAGE_SIZE };
> 
> Why not sizeof(shared_info), utilizing the zero padding on the
> receiving side?
> 

Ok, yes, I guess that would work.

> > +    size_t hdr_size = offsetof(typeof(ctxt), buffer);
> > +    int rc;
> > +
> > +    rc = DOMAIN_SAVE_BEGIN(SHARED_INFO, c, 0);
> > +    if ( rc )
> > +        return rc;
> > +
> > +#ifdef CONFIG_COMPAT
> > +    if ( !dry_run )
> > +        ctxt.has_32bit_shinfo = has_32bit_shinfo(d);
> > +#endif
> 
> Nothing will go wrong without the if(), I suppose? Better drop it
> then? It could then also easily be part of the initializer of ctxt.
> 

Ok. I said last time I wanted to keep it as it was illustrative but I'll drop 
it since it has now come up twice.

> > +    rc = domain_save_data(c, &ctxt, hdr_size);
> > +    if ( rc )
> > +        return rc;
> > +
> > +    rc = domain_save_data(c, d->shared_info, ctxt.buffer_size);
> > +    if ( rc )
> > +        return rc;
> > +
> > +    return domain_save_end(c);
> > +}
> > +
> > +static int load_shared_info(struct domain *d, struct domain_context *c)
> > +{
> > +    struct domain_shared_info_context ctxt;
> > +    size_t hdr_size = offsetof(typeof(ctxt), buffer);
> > +    unsigned int i;
> > +    int rc;
> > +
> > +    rc = DOMAIN_LOAD_BEGIN(SHARED_INFO, c, &i);
> > +    if ( rc || i ) /* expect only a single instance */
> > +        return rc;
> > +
> > +    rc = domain_load_data(c, &ctxt, hdr_size);
> > +    if ( rc )
> > +        return rc;
> > +
> > +    if ( ctxt.pad[0] || ctxt.pad[1] || ctxt.pad[2] ||
> > +         ctxt.buffer_size != PAGE_SIZE )
> > +        return -EINVAL;
> > +
> > +#ifdef CONFIG_COMPAT
> > +    d->arch.has_32bit_shinfo = ctxt.has_32bit_shinfo;
> > +#endif
> 
> There's nothing wrong with using has_32bit_shinfo(d) here as well.
> 

I just thought it looked odd.

> > --- a/xen/include/public/save.h
> > +++ b/xen/include/public/save.h
> > @@ -73,7 +73,16 @@ struct domain_save_header {
> >  };
> >  DECLARE_DOMAIN_SAVE_TYPE(HEADER, 1, struct domain_save_header);
> >
> > -#define DOMAIN_SAVE_CODE_MAX 1
> > +struct domain_shared_info_context {
> > +    uint8_t has_32bit_shinfo;
> > +    uint8_t pad[3];
> 
> 32-(or 16-)bit flags, with just a single bit used for the purpose?
> 

I debated that. Given this is xen/tools-only would a bit-field be acceptable?

  Paul

> Jan




 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.