|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] Re: [PATCH V4 03/15] x86/pv: Rewrite how building PV dom0 handles domheap mappings
Hi,
I've been trying to run this series for a while, but it crashes very
frequentyly starting from the patch that generalizes the mapcache. I think I've
tracked it down to this patch.
On Mon Nov 11, 2024 at 1:11 PM GMT, Elias El Yandouzi wrote:
> From: Hongyan Xia <hongyxia@xxxxxxxxxx>
>
> Building a PV dom0 is allocating from the domheap but uses it like the
> xenheap. Use the pages as they should be.
>
> Signed-off-by: Hongyan Xia <hongyxia@xxxxxxxxxx>
> Signed-off-by: Julien Grall <jgrall@xxxxxxxxxx>
> Signed-off-by: Elias El Yandouzi <eliasely@xxxxxxxxxx>
>
> ----
> Changes in V4:
> * Reduce the scope of l{1,2,4}start_mfn variables
> * Make the macro `UNMAP_MAP_AND_ADVANCE` return the new virtual
> address
>
> Changes in V3:
> * Fold following patch 'x86/pv: Map L4 page table for shim domain'
>
> Changes in V2:
> * Clarify the commit message
> * Break the patch in two parts
>
> Changes since Hongyan's version:
> * Rebase
> * Remove spurious newline
>
> diff --git a/xen/arch/x86/pv/dom0_build.c b/xen/arch/x86/pv/dom0_build.c
> index 18b7a3e4e025..b03df609cadb 100644
> --- a/xen/arch/x86/pv/dom0_build.c
> +++ b/xen/arch/x86/pv/dom0_build.c
> @@ -382,6 +382,7 @@ static int __init dom0_construct(struct domain *d,
> l3_pgentry_t *l3tab = NULL, *l3start = NULL;
> l2_pgentry_t *l2tab = NULL, *l2start = NULL;
> l1_pgentry_t *l1tab = NULL, *l1start = NULL;
> + mfn_t l3start_mfn = INVALID_MFN;
>
> /*
> * This fully describes the memory layout of the initial domain. All
> @@ -719,22 +720,34 @@ static int __init dom0_construct(struct domain *d,
> v->arch.pv.event_callback_cs = FLAT_COMPAT_KERNEL_CS;
> }
>
> +#define UNMAP_MAP_AND_ADVANCE(mfn_var, virt_var, maddr) ({ \
> + do { \
> + unmap_domain_page(virt_var); \
> + mfn_var = maddr_to_mfn(maddr); \
> + maddr += PAGE_SIZE; \
> + virt_var = map_domain_page(mfn_var); \
> + } while ( false ); \
> + virt_var; \
> +})
> +
> if ( !compat )
> {
> + mfn_t l4start_mfn;
> maddr_to_page(mpt_alloc)->u.inuse.type_info = PGT_l4_page_table;
> - l4start = l4tab = __va(mpt_alloc); mpt_alloc += PAGE_SIZE;
> + l4tab = UNMAP_MAP_AND_ADVANCE(l4start_mfn, l4start, mpt_alloc);
In here l4start is mapped on the idle domain perdomain area, but...
> clear_page(l4tab);
> - init_xen_l4_slots(l4tab, _mfn(virt_to_mfn(l4start)),
> - d, INVALID_MFN, true);
> - v->arch.guest_table = pagetable_from_paddr(__pa(l4start));
> + init_xen_l4_slots(l4tab, l4start_mfn, d, INVALID_MFN, true);
> + v->arch.guest_table = pagetable_from_mfn(l4start_mfn);
> }
> else
> {
> /* Monitor table already created by switch_compat(). */
> - l4start = l4tab = __va(pagetable_get_paddr(v->arch.guest_table));
> + mfn_t l4start_mfn = pagetable_get_mfn(v->arch.guest_table);
> + l4start = l4tab = map_domain_page(l4start_mfn);
> /* See public/xen.h on why the following is needed. */
> maddr_to_page(mpt_alloc)->u.inuse.type_info = PGT_l3_page_table;
> l3start = __va(mpt_alloc); mpt_alloc += PAGE_SIZE;
> + UNMAP_MAP_AND_ADVANCE(l3start_mfn, l3start, mpt_alloc);
> }
>
> l4tab += l4_table_offset(v_start);
> @@ -743,15 +756,17 @@ static int __init dom0_construct(struct domain *d,
> {
> if ( !((unsigned long)l1tab & (PAGE_SIZE-1)) )
> {
> + mfn_t l1start_mfn;
> maddr_to_page(mpt_alloc)->u.inuse.type_info = PGT_l1_page_table;
> - l1start = l1tab = __va(mpt_alloc); mpt_alloc += PAGE_SIZE;
> + l1tab = UNMAP_MAP_AND_ADVANCE(l1start_mfn, l1start, mpt_alloc);
> clear_page(l1tab);
> if ( count == 0 )
> l1tab += l1_table_offset(v_start);
> if ( !((unsigned long)l2tab & (PAGE_SIZE-1)) )
> {
> + mfn_t l2start_mfn;
> maddr_to_page(mpt_alloc)->u.inuse.type_info =
> PGT_l2_page_table;
> - l2start = l2tab = __va(mpt_alloc); mpt_alloc += PAGE_SIZE;
> + l2tab = UNMAP_MAP_AND_ADVANCE(l2start_mfn, l2start,
> mpt_alloc);
> clear_page(l2tab);
> if ( count == 0 )
> l2tab += l2_table_offset(v_start);
> @@ -761,19 +776,19 @@ static int __init dom0_construct(struct domain *d,
> {
> maddr_to_page(mpt_alloc)->u.inuse.type_info =
> PGT_l3_page_table;
> - l3start = __va(mpt_alloc); mpt_alloc += PAGE_SIZE;
> + UNMAP_MAP_AND_ADVANCE(l3start_mfn, l3start,
> mpt_alloc);
> }
> l3tab = l3start;
> clear_page(l3tab);
> if ( count == 0 )
> l3tab += l3_table_offset(v_start);
> - *l4tab = l4e_from_paddr(__pa(l3start), L4_PROT);
> + *l4tab = l4e_from_mfn(l3start_mfn, L4_PROT);
> l4tab++;
> }
> - *l3tab = l3e_from_paddr(__pa(l2start), L3_PROT);
> + *l3tab = l3e_from_mfn(l2start_mfn, L3_PROT);
> l3tab++;
> }
> - *l2tab = l2e_from_paddr(__pa(l1start), L2_PROT);
> + *l2tab = l2e_from_mfn(l1start_mfn, L2_PROT);
> l2tab++;
> }
> if ( count < initrd_pfn || count >= initrd_pfn + PFN_UP(initrd_len) )
> @@ -792,27 +807,32 @@ static int __init dom0_construct(struct domain *d,
>
> if ( compat )
> {
> - l2_pgentry_t *l2t;
> -
> /* Ensure the first four L3 entries are all populated. */
> for ( i = 0, l3tab = l3start; i < 4; ++i, ++l3tab )
> {
> if ( !l3e_get_intpte(*l3tab) )
> {
> + mfn_t l2start_mfn;
> maddr_to_page(mpt_alloc)->u.inuse.type_info =
> PGT_l2_page_table;
> - l2tab = __va(mpt_alloc); mpt_alloc += PAGE_SIZE;
> - clear_page(l2tab);
> - *l3tab = l3e_from_paddr(__pa(l2tab), L3_PROT);
> + UNMAP_MAP_AND_ADVANCE(l2start_mfn, l2start, mpt_alloc);
> + clear_page(l2start);
> + *l3tab = l3e_from_mfn(l2start_mfn, L3_PROT);
> }
> if ( i == 3 )
> l3e_get_page(*l3tab)->u.inuse.type_info |= PGT_pae_xen_l2;
> }
>
> - l2t = map_l2t_from_l3e(l3start[3]);
> - init_xen_pae_l2_slots(l2t, d);
> - unmap_domain_page(l2t);
> + UNMAP_DOMAIN_PAGE(l2start);
> + l2start = map_l2t_from_l3e(l3start[3]);
> + init_xen_pae_l2_slots(l2start, d);
> }
>
> +#undef UNMAP_MAP_AND_ADVANCE
> +
> + UNMAP_DOMAIN_PAGE(l1start);
> + UNMAP_DOMAIN_PAGE(l2start);
> + UNMAP_DOMAIN_PAGE(l3start);
... l4start is not unmapped here. This is a problem, because we're about to
change the page tables into dom0's and start using its mapcache.
IMO, we should be unmapping here, and remapping in dom0's context. Otherwise
l4start becomes a transiently stale pointer. Any remaining pointer obtained via
map_domain_page() is a dangling pointer after the mapcache+pagetable switch.
> +
> /* Pages that are part of page tables must be read only. */
> mark_pv_pt_pages_rdonly(d, l4start, vpt_start, nr_pt_pages,
> &flush_flags);
>
> @@ -987,6 +1007,8 @@ static int __init dom0_construct(struct domain *d,
> pv_shim_setup_dom(d, l4start, v_start, vxenstore_start,
> vconsole_start,
> vphysmap_start, si);
>
> + UNMAP_DOMAIN_PAGE(l4start);
As it is, this unmap is operating on the wrong mapcache, I think. I don't quite
understand why I see intermittent boot crashes and not constant ones, but this
seems like a bug.
What we want, I think, is:
1. Increase the scope of l4start_mfn to be function-level.
2. Do UNMAP_DOMAIN_PAGE(l4start) along with l1start, l2start and l3start.
3. Include a pair of map_domain_page() and UNMAP_DOMAIN_PAGE() within the
conditional, surrounding pv_shim_setup_dom.
> +
> #ifdef CONFIG_COMPAT
> if ( compat )
> xlat_start_info(si, pv_shim ? XLAT_start_info_console_domU
I'll keep testing it in case I missed something, but this seems to work.
Cheers,
Alejandro
|
 |
Lists.xenproject.org is hosted with RackSpace, monitoring our |