Re: [Xen-devel] [PATCH v2] libxc: add xc_domain_add_to_physmap_batch to wrap XENMEM_add_to_physmap_batch
Hi Wei, 2017-06-21 23:44 GMT+08:00 Wei Liu: > On Wed, Jun 21, 2017 at 01:29:26AM +0800, Zhongze Liu wrote: >> This is a preparation for the proposal "allow setting up shared memory areas >> between VMs from xl config file". See: >> V2: https://lists.xen.org/archives/html/xen-devel/2017-06/msg02256.html >> V1: https://lists.xen.org/archives/html/xen-devel/2017-05/msg01288.html >> >> The plan is to use XENMEM_add_to_physmap_batch in xl to map foregin pages >> from >> one DomU to another so that the page could be shared. But currently there is >> no >> wrapper for XENMEM_add_to_physmap_batch in libxc, so we just add a wrapper >> for >> it. >> >> Signed-off-by: Zhongze Liu >> --- >> +int xc_domain_add_to_physmap_batch(xc_interface *xch, >> + domid_t domid, >> + domid_t foreign_domid, >> + unsigned int space, >> + unsigned int size, >> + xen_ulong_t *idxs, >> + xen_pfn_t *gpfns, >> + . int *errs) >> +{ >> +int rc; >> +DECLARE_HYPERCALL_BOUNCE(idxs, size * sizeof(*idxs), >> XC_HYPERCALL_BUFFER_BOUNCE_IN); >> +DECLARE_HYPERCALL_BOUNCE(gpfns, size * sizeof(*gpfns), >> XC_HYPERCALL_BUFFER_BOUNCE_IN); >> +DECLARE_HYPERCALL_BOUNCE(errs, size * sizeof(*errs), >> XC_HYPERCALL_BUFFER_BOUNCE_OUT); >> + >> +struct xen_add_to_physmap_batch xatp_batch = { >> +.domid = domid, >> +.space = space, >> +.size = size, >> +.u = {.foreign_domid = foreign_domid} > > Coding style issue. Do you mean that I should add a space between '{' and '.' near ".u = {.foreign" in this line? > > Just a note, the struct is different for pre-4.7 and post-4.7 Xen. You > don't need to implement a version of this function for pre-4.7 Xen. > >> +}; >> + >> +if ( xc_hypercall_bounce_pre(xch, idxs) || >> + xc_hypercall_bounce_pre(xch, gpfns) || >> + xc_hypercall_bounce_pre(xch, errs) ) >> +{ >> +PERROR("Could not bounce memory for XENMEM_add_to_physmap_batch"); >> +goto out; > > rc will be uninitialised in this exit path. > >> +} >> + >> +set_xen_guest_handle(xatp_batch.idxs, idxs); >> +set_xen_guest_handle(xatp_batch.gpfns, gpfns); >> +set_xen_guest_handle(xatp_batch.errs, errs); >> + >> +rc = do_memory_op(xch, XENMEM_add_to_physmap_batch, >> + _batch, sizeof(xatp_batch)); >> + >> +out: >> +xc_hypercall_bounce_post(xch, idxs); >> +xc_hypercall_bounce_post(xch, gpfns); >> +xc_hypercall_bounce_post(xch, errs); >> + >> +return rc; >> +} >> + >> int xc_domain_claim_pages(xc_interface *xch, >> uint32_t domid, >> unsigned long nr_pages) >> -- >> 2.13.1 >> Cheers, Zhongze Liu ___ Xen-devel mailing list Xen-devel@lists.xen.org https://lists.xen.org/xen-devel
Re: [Xen-devel] [PATCH v2] libxc: add xc_domain_add_to_physmap_batch to wrap XENMEM_add_to_physmap_batch
On Wed, Jun 21, 2017 at 01:29:26AM +0800, Zhongze Liu wrote: > This is a preparation for the proposal "allow setting up shared memory areas > between VMs from xl config file". See: > V2: https://lists.xen.org/archives/html/xen-devel/2017-06/msg02256.html > V1: https://lists.xen.org/archives/html/xen-devel/2017-05/msg01288.html > > The plan is to use XENMEM_add_to_physmap_batch in xl to map foregin pages from > one DomU to another so that the page could be shared. But currently there is > no > wrapper for XENMEM_add_to_physmap_batch in libxc, so we just add a wrapper for > it. > > Signed-off-by: Zhongze Liu> --- > +int xc_domain_add_to_physmap_batch(xc_interface *xch, > + domid_t domid, > + domid_t foreign_domid, > + unsigned int space, > + unsigned int size, > + xen_ulong_t *idxs, > + xen_pfn_t *gpfns, > + int *errs) > +{ > +int rc; > +DECLARE_HYPERCALL_BOUNCE(idxs, size * sizeof(*idxs), > XC_HYPERCALL_BUFFER_BOUNCE_IN); > +DECLARE_HYPERCALL_BOUNCE(gpfns, size * sizeof(*gpfns), > XC_HYPERCALL_BUFFER_BOUNCE_IN); > +DECLARE_HYPERCALL_BOUNCE(errs, size * sizeof(*errs), > XC_HYPERCALL_BUFFER_BOUNCE_OUT); > + > +struct xen_add_to_physmap_batch xatp_batch = { > +.domid = domid, > +.space = space, > +.size = size, > +.u = {.foreign_domid = foreign_domid} Coding style issue. Just a note, the struct is different for pre-4.7 and post-4.7 Xen. You don't need to implement a version of this function for pre-4.7 Xen. > +}; > + > +if ( xc_hypercall_bounce_pre(xch, idxs) || > + xc_hypercall_bounce_pre(xch, gpfns) || > + xc_hypercall_bounce_pre(xch, errs) ) > +{ > +PERROR("Could not bounce memory for XENMEM_add_to_physmap_batch"); > +goto out; rc will be uninitialised in this exit path. > +} > + > +set_xen_guest_handle(xatp_batch.idxs, idxs); > +set_xen_guest_handle(xatp_batch.gpfns, gpfns); > +set_xen_guest_handle(xatp_batch.errs, errs); > + > +rc = do_memory_op(xch, XENMEM_add_to_physmap_batch, > + _batch, sizeof(xatp_batch)); > + > +out: > +xc_hypercall_bounce_post(xch, idxs); > +xc_hypercall_bounce_post(xch, gpfns); > +xc_hypercall_bounce_post(xch, errs); > + > +return rc; > +} > + > int xc_domain_claim_pages(xc_interface *xch, > uint32_t domid, > unsigned long nr_pages) > -- > 2.13.1 > ___ Xen-devel mailing list Xen-devel@lists.xen.org https://lists.xen.org/xen-devel
[Xen-devel] [PATCH v2] libxc: add xc_domain_add_to_physmap_batch to wrap XENMEM_add_to_physmap_batch
This is a preparation for the proposal "allow setting up shared memory areas between VMs from xl config file". See: V2: https://lists.xen.org/archives/html/xen-devel/2017-06/msg02256.html V1: https://lists.xen.org/archives/html/xen-devel/2017-05/msg01288.html The plan is to use XENMEM_add_to_physmap_batch in xl to map foregin pages from one DomU to another so that the page could be shared. But currently there is no wrapper for XENMEM_add_to_physmap_batch in libxc, so we just add a wrapper for it. Signed-off-by: Zhongze Liu--- Changed Since v1: * explain why such a sudden wrapper * change the parameters' types Cc: Ian Jackson , Cc: Wei Liu , Cc: Stefano Stabellini Cc: Julien Grall Cc: Jan Beulich