Re: [PATCH V10 05/14] xen/pvh: balloon and grant changes.

From: Konrad Rzeszutek Wilk
Date: Fri Dec 13 2013 - 09:37:37 EST


On Fri, Dec 13, 2013 at 11:56:31AM +0000, Stefano Stabellini wrote:
> On Thu, 12 Dec 2013, Konrad Rzeszutek Wilk wrote:
> > From: Mukesh Rathor <mukesh.rathor@xxxxxxxxxx>
> >
> > For balloon changes we skip setting of local P2M as it's updated
> > in Xen. For grant, the shared grant frame is the PFN and not MFN,
> > hence its mapped via the same code path as HVM.
> >
> > Signed-off-by: Mukesh Rathor <mukesh.rathor@xxxxxxxxxx>
> > [v1: Rebase on v3.7-rc3]
> > Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>
> > CC: <stefano.stabellini@xxxxxxxxxxxxx>
> > Conflicts:
> > drivers/xen/balloon.c
> > [v2: Rebase on v3.13-rc3 has lot of ARM changes.]
> > ---
> > drivers/xen/balloon.c | 37 ++++++++++++++++++++-----------------
> > drivers/xen/gntdev.c | 3 ++-
> > drivers/xen/grant-table.c | 25 ++++++++++++++++++++++---
> > 3 files changed, 44 insertions(+), 21 deletions(-)
> >
> > diff --git a/drivers/xen/balloon.c b/drivers/xen/balloon.c
> > index 55ea73f..0e9875a 100644
> > --- a/drivers/xen/balloon.c
> > +++ b/drivers/xen/balloon.c
> > @@ -354,7 +354,9 @@ static enum bp_state increase_reservation(unsigned long nr_pages)
> >
> > #ifdef CONFIG_XEN_HAVE_PVMMU
> > /* Link back into the page tables if not highmem. */
> > - if (xen_pv_domain() && !PageHighMem(page)) {
> > + if (xen_pv_domain() && !PageHighMem(page) &&
> > + !xen_feature(XENFEAT_auto_translated_physmap)) {
> > +
> > int ret;
> > ret = HYPERVISOR_update_va_mapping(
> > (unsigned long)__va(pfn << PAGE_SHIFT),
> > @@ -411,28 +413,29 @@ static enum bp_state decrease_reservation(unsigned long nr_pages, gfp_t gfp)
> >
> > scrub_page(page);
> >
> > - /*
> > - * Ballooned out frames are effectively replaced with
> > - * a scratch frame. Ensure direct mappings and the
> > - * p2m are consistent.
> > - */
> > - scratch_page = get_balloon_scratch_page();
> > + if (!xen_feature(XENFEAT_auto_translated_physmap)) {
> > + unsigned long p;
> > +
> > + /*
> > + * Ballooned out frames are effectively replaced with
> > + * a scratch frame. Ensure direct mappings and the
> > + * p2m are consistent.
> > + */
> > + scratch_page = get_balloon_scratch_page();
> > #ifdef CONFIG_XEN_HAVE_PVMMU
> > - if (xen_pv_domain() && !PageHighMem(page)) {
> > - ret = HYPERVISOR_update_va_mapping(
> > - (unsigned long)__va(pfn << PAGE_SHIFT),
> > - pfn_pte(page_to_pfn(scratch_page),
> > + if (xen_pv_domain() && !PageHighMem(page)) {
> > + ret = HYPERVISOR_update_va_mapping(
> > + (unsigned long)__va(pfn << PAGE_SHIFT),
> > + pfn_pte(page_to_pfn(scratch_page),
> > PAGE_KERNEL_RO), 0);
> > - BUG_ON(ret);
> > - }
> > + BUG_ON(ret);
> > + }
> > #endif
> > - if (!xen_feature(XENFEAT_auto_translated_physmap)) {
> > - unsigned long p;
> > p = page_to_pfn(scratch_page);
> > __set_phys_to_machine(pfn, pfn_to_mfn(p));
> > - }
> > - put_balloon_scratch_page();
> >
> > + put_balloon_scratch_page();
> > + }
> > balloon_append(pfn_to_page(pfn));
> > }
>
> These changes to balloon.c are similar to my patch but:
>
> - they don't prevent set_phys_to_machine from being called by
> increase_reservation for auto_translated guests

<nods>
>
> - they don't prevent wasteful balloon_scratch_page from being allocated
> for auto_translated guests

<nods>
>
> I think that it would be best to take my patch instead.

Lets focus the review on your patch then and have it part of this
series.
>
>
>
> > diff --git a/drivers/xen/gntdev.c b/drivers/xen/gntdev.c
> > index e41c79c..c1ae3ce 100644
> > --- a/drivers/xen/gntdev.c
> > +++ b/drivers/xen/gntdev.c
> > @@ -846,7 +846,8 @@ static int __init gntdev_init(void)
> > if (!xen_domain())
> > return -ENODEV;
> >
> > - use_ptemod = xen_pv_domain();
> > + use_ptemod = xen_pv_domain() &&
> > + !xen_feature(XENFEAT_auto_translated_physmap);
> >
> > err = misc_register(&gntdev_miscdev);
> > if (err != 0) {
>
> you might actually just want to have:
>
> use_ptemod = !xen_feature(XENFEAT_auto_translated_physmap);
>
>
>
>
> > diff --git a/drivers/xen/grant-table.c b/drivers/xen/grant-table.c
> > index 0283871..1fcf75b 100644
> > --- a/drivers/xen/grant-table.c
> > +++ b/drivers/xen/grant-table.c
> > @@ -1056,14 +1056,20 @@ static void gnttab_unmap_frames_v2(void)
> > static int gnttab_map(unsigned int start_idx, unsigned int end_idx)
> > {
> > struct gnttab_setup_table setup;
> > + unsigned long start_gpfn;
> > xen_pfn_t *frames;
> > unsigned int nr_gframes = end_idx + 1;
> > int rc;
> >
> > - if (xen_hvm_domain()) {
> > + if (xen_hvm_domain() || xen_feature(XENFEAT_auto_translated_physmap)) {
> > struct xen_add_to_physmap xatp;
> > unsigned int i = end_idx;
> > rc = 0;
> > +
> > + if (xen_hvm_domain())
> > + start_gpfn = xen_hvm_resume_frames >> PAGE_SHIFT;
> > + else
> > + start_gpfn = virt_to_pfn(gnttab_shared.addr);
> > /*
> > * Loop backwards, so that the first hypercall has the largest
> > * index, ensuring that the table will grow only once.
> > @@ -1072,7 +1078,7 @@ static int gnttab_map(unsigned int start_idx, unsigned int end_idx)
> > xatp.domid = DOMID_SELF;
> > xatp.idx = i;
> > xatp.space = XENMAPSPACE_grant_table;
> > - xatp.gpfn = (xen_hvm_resume_frames >> PAGE_SHIFT) + i;
> > + xatp.gpfn = start_gpfn + i;
> > rc = HYPERVISOR_memory_op(XENMEM_add_to_physmap, &xatp);
> > if (rc != 0) {
> > pr_warn("grant table add_to_physmap failed, err=%d\n",
> > @@ -1135,7 +1141,7 @@ static void gnttab_request_version(void)
> > int rc;
> > struct gnttab_set_version gsv;
> >
> > - if (xen_hvm_domain())
> > + if (xen_hvm_domain() || xen_feature(XENFEAT_auto_translated_physmap))
> > gsv.version = 1;
> > else
> > gsv.version = 2;
> > @@ -1164,11 +1170,24 @@ static void gnttab_request_version(void)
> > static int gnttab_setup(void)
> > {
> > unsigned int max_nr_gframes;
> > + char *kmsg = "Failed to kmalloc pages for pv in hvm grant frames\n";
> >
> > max_nr_gframes = gnttab_max_grant_frames();
> > if (max_nr_gframes < nr_grant_frames)
> > return -ENOSYS;
> >
> > + /* PVH note: xen will free existing kmalloc'd mfn in
> > + * XENMEM_add_to_physmap. TBD/FIXME: use xen ballooning instead of
> > + * kmalloc(). */
> > + if (xen_pv_domain() && xen_feature(XENFEAT_auto_translated_physmap) &&
> > + !gnttab_shared.addr) {
> > + gnttab_shared.addr =
> > + kmalloc(max_nr_gframes * PAGE_SIZE, GFP_KERNEL);
> > + if (!gnttab_shared.addr) {
> > + pr_warn("%s", kmsg);
> > + return -ENOMEM;
> > + }
> > + }
> > if (xen_pv_domain())
> > return gnttab_map(0, nr_grant_frames - 1);
>
> Can't you just use xen_hvm_resume_frames instead? Just rename it
> xen_auto_translated_frames and allocate memory to it from
> arch/x86/xen/enlighten.c.
> Then you can completely unify the hvm and pvh code paths in this file,
> as a result all the
>
> if (xen_hvm_domain() || xen_feature(XENFEAT_auto_translated_physmap))
>
> tests, could be reduced to
>
> if (xen_feature(XENFEAT_auto_translated_physmap))

Mukesh? That should work right?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/