Hi Luca, On 15/05/2024 16:26, Luca Fancellu wrote: > > > From: Penny Zheng <penny.zh...@arm.com> > > We are doing foreign memory mapping for static shared memory, and > there is a great possibility that it could be super mapped. > But today, p2m_put_l3_page could not handle superpages. s/could/can
> > This commits implements a new function p2m_put_l2_superpage to handle > 2MB superpages, specifically for helping put extra references for > foreign superpages. > > Modify relinquish_p2m_mapping as well to take into account preemption > when type is foreign memory and order is above 9 (2MB). I don't see the reasoning why we don't handle 1GB super pages. It would be beneficial to include such in commit msg as well as in the code. > > Signed-off-by: Penny Zheng <penny.zh...@arm.com> > Signed-off-by: Luca Fancellu <luca.fance...@arm.com> > --- > v2: > - Do not handle 1GB super page as there might be some issue where > a lot of calls to put_page(...) might be issued which could lead > to free memory that is a long operation. > v1: > - patch from > https://patchwork.kernel.org/project/xen-devel/patch/20231206090623.1932275-9-penny.zh...@arm.com/ > --- > xen/arch/arm/mmu/p2m.c | 57 +++++++++++++++++++++++++++++------------- > 1 file changed, 40 insertions(+), 17 deletions(-) > > diff --git a/xen/arch/arm/mmu/p2m.c b/xen/arch/arm/mmu/p2m.c > index 41fcca011cf4..29fdb3b87dd0 100644 > --- a/xen/arch/arm/mmu/p2m.c > +++ b/xen/arch/arm/mmu/p2m.c > @@ -753,17 +753,9 @@ static int p2m_mem_access_radix_set(struct p2m_domain > *p2m, gfn_t gfn, > return rc; > } > > -/* > - * Put any references on the single 4K page referenced by pte. > - * TODO: Handle superpages, for now we only take special references for leaf > - * pages (specifically foreign ones, which can't be super mapped today). > - */ > -static void p2m_put_l3_page(const lpae_t pte) > +/* Put any references on the single 4K page referenced by mfn. */ > +static void p2m_put_l3_page(mfn_t mfn, p2m_type_t type) > { > - mfn_t mfn = lpae_get_mfn(pte); > - > - ASSERT(p2m_is_valid(pte)); > - > /* > * TODO: Handle other p2m types > * > @@ -771,16 +763,44 @@ static void p2m_put_l3_page(const lpae_t pte) > * flush the TLBs if the page is reallocated before the end of > * this loop. > */ > - if ( p2m_is_foreign(pte.p2m.type) ) > + if ( p2m_is_foreign(type) ) > { > ASSERT(mfn_valid(mfn)); > put_page(mfn_to_page(mfn)); > } > /* Detect the xenheap page and mark the stored GFN as invalid. */ > - else if ( p2m_is_ram(pte.p2m.type) && is_xen_heap_mfn(mfn) ) > + else if ( p2m_is_ram(type) && is_xen_heap_mfn(mfn) ) > page_set_xenheap_gfn(mfn_to_page(mfn), INVALID_GFN); > } > > +/* Put any references on the superpage referenced by mfn. */ > +static void p2m_put_l2_superpage(mfn_t mfn, p2m_type_t type) > +{ > + unsigned int i; > + unsigned int level_order = XEN_PT_LEVEL_ORDER(3); If we know the third level order is always 0 no matter the page shift, does it make sense to have this variable? > + > + for ( i = 0; i < XEN_PT_LPAE_ENTRIES; i++ ) > + { > + p2m_put_l3_page(mfn, type); > + > + mfn = mfn_add(mfn, 1 << level_order); > + } > +} > + > +/* Put any references on the page referenced by pte. */ > +static void p2m_put_page(const lpae_t pte, unsigned int level) > +{ > + mfn_t mfn = lpae_get_mfn(pte); > + > + ASSERT(p2m_is_valid(pte)); > + > + /* We have a second level 2M superpage */ > + if ( p2m_is_superpage(pte, level) && (level == 2) ) > + return p2m_put_l2_superpage(mfn, pte.p2m.type); > + else if ( level == 3 ) > + return p2m_put_l3_page(mfn, pte.p2m.type); > +} > + > /* Free lpae sub-tree behind an entry */ > static void p2m_free_entry(struct p2m_domain *p2m, > lpae_t entry, unsigned int level) > @@ -809,9 +829,10 @@ static void p2m_free_entry(struct p2m_domain *p2m, > #endif > > p2m->stats.mappings[level]--; > - /* Nothing to do if the entry is a super-page. */ > - if ( level == 3 ) > - p2m_put_l3_page(entry); > + /* TODO: Currently we don't handle 1GB super-page. */ As a future reader of the code it would be beneficial to say why. > + if ( level >= 2 ) > + p2m_put_page(entry, level); > + > return; > } > > @@ -1558,9 +1579,11 @@ int relinquish_p2m_mapping(struct domain *d) > > count++; > /* > - * Arbitrarily preempt every 512 iterations. > + * Arbitrarily preempt every 512 iterations or when type is foreign > + * mapping and the order is above 9 (2MB). > */ > - if ( !(count % 512) && hypercall_preempt_check() ) > + if ( (!(count % 512) || (p2m_is_foreign(t) && (order > 9))) && Instead of (order > 9), use XEN_PT_LEVEL_ORDER(2) > + hypercall_preempt_check() ) > { > rc = -ERESTART; > break; > -- > 2.34.1 > ~Michal
