On Wed, 2015-11-11 at 15:13 +0000, Ian Jackson wrote:
> Ian Campbell writes ("[PATCH XEN v5 13/23] tools: Refactor foreign memory
> mapping into libxenforeignmemory"):
> > libxenforeignmemory will provide a stable API and ABI for mapping
> > foreign domain memory (subject to appropriate privileges).
> >
> > The new library exposes an interface equivalent to
> > xc_map_foreign_memory_bulk, which all the other
> > xc_map_foreign_memory_* functions (which remain in libxc) are
> > implemented in terms of.
> ...
> > diff --git a/tools/libs/foreignmemory/include/xenforeignmemory.h
> > b/tools/libs/foreignmemory/include/xenforeignmemory.h
>
> > +/*
> > + * Return a handle onto the hypercall driver. Logs errors.
> > + */
> > +xenforeignmemory_handle *xenforeignmemory_open(xentoollog_logger
> > *logger,
> > + unsigned open_flags);
>
> `onto the foreign memory driver'.
>
> This file should probably say something about fork(). Perhaps just a
> refernce to the rules in the grant map driver.
Yes, same rules I think.
>
> > +/*
> > + * Maps a range within one domain to a local address range. Mappings
> > + * should be unmapped with munmap and should follow the same rules as
> > mmap
> > + * regarding page alignment.
> > + *
> > + * prot is as for mmap(2).
> > + *
> > + * Can partially succeed. When a page cannot be mapped, its respective
> > + * field in @err is set to the corresponding errno value.
> > + *
> > + * Returns NULL if no pages can be mapped.
> ...
> > +void *xenforeignmemory_map(xenforeignmemory_handle *fmem, uint32_t
> > dom,
> > + int prot, const xen_pfn_t *arr, int *err,
> > + unsigned int num);
>
> If it returns NULL, will all of *err be filled with errno values ?
> Ie is all of err[] always written ?
>
> Does this function ever set errno ?
I'll figure all these out and say.
> I would write `int err_out[num]' in the prototype which is a bit
> clearer about the semantics.
Are you suggesting:
void *xenforeignmemory_map(xenforeignmemory_handle *fmem, uint32_t dom,
int prot, const xen_pfn_t *arr,
int err_out[num],
unsigned int num);
Is that (a var sized array based on another param) really allowed?
Or did you mean as comment like:
void *xenforeignmemory_map(xenforeignmemory_handle *fmem, uint32_t dom,
int prot, const xen_pfn_t *arr,
int *err_out /*int err_out[num]*/,
unsigned
int num);
> This API invites callers to fail to notice partial failures.
>
> How about permitting err==NULL to mean `on partial failure, unmap
> everything and return NULL setting errno' ?
Not a bad idea.
Ian.
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xen.org
http://lists.xen.org/xen-devel
