| Message ID | 20190218113600.9540-8-julien.grall@arm.com |
|---|---|
| State | Superseded |
| Headers | show |
| Series | xen/arm: Properly disable M2P on Arm. | expand |
>>> On 18.02.19 at 12:35, <julien.grall@arm.com> wrote: > --- a/xen/arch/x86/mm.c > +++ b/xen/arch/x86/mm.c > @@ -4300,7 +4300,8 @@ int xenmem_add_to_physmap_one( > { > struct page_info *page = NULL; > unsigned long gfn = 0; /* gcc ... */ > - unsigned long prev_mfn, old_gpfn; > + mfn_t prev_mfn; > + unsigned long old_gpfn; Please can you put this together with "gfn"? > --- a/xen/common/page_alloc.c > +++ b/xen/common/page_alloc.c > @@ -2121,9 +2121,9 @@ void init_xenheap_pages(paddr_t ps, paddr_t pe) > * Yuk! Ensure there is a one-page buffer between Xen and Dom zones, to > * prevent merging of power-of-two blocks across the zone boundary. > */ > - if ( ps && !is_xen_heap_mfn(paddr_to_pfn(ps)-1) ) > + if ( ps && !is_xen_heap_mfn(_mfn(paddr_to_pfn(ps)-1)) ) Please add the missing blanks around - at the same time. > ps += PAGE_SIZE; > - if ( !is_xen_heap_mfn(paddr_to_pfn(pe)) ) > + if ( !is_xen_heap_mfn(maddr_to_mfn(pe)) ) Why maddr_to_mfn() here but still paddr_to_pfn() above? Oh, we don't have any mfn_sub(), I see. > --- a/xen/include/asm-x86/mm.h > +++ b/xen/include/asm-x86/mm.h > @@ -278,7 +278,7 @@ struct page_info > > #define is_xen_heap_page(page) ((page)->count_info & PGC_xen_heap) > #define is_xen_heap_mfn(mfn) \ > - (__mfn_valid(mfn) && is_xen_heap_page(mfn_to_page(_mfn(mfn)))) > + (__mfn_valid(mfn_x(mfn)) && is_xen_heap_page(mfn_to_page(mfn))) Please don't open code mfn_valid(). With these minor issues taken care of Reviewed-by: Jan Beulich <jbeulich@suse.com> Jan
Hi, On 13/03/2019 15:04, Jan Beulich wrote: >>>> On 18.02.19 at 12:35, <julien.grall@arm.com> wrote: >> --- a/xen/arch/x86/mm.c >> +++ b/xen/arch/x86/mm.c >> @@ -4300,7 +4300,8 @@ int xenmem_add_to_physmap_one( >> { >> struct page_info *page = NULL; >> unsigned long gfn = 0; /* gcc ... */ >> - unsigned long prev_mfn, old_gpfn; >> + mfn_t prev_mfn; >> + unsigned long old_gpfn; > > Please can you put this together with "gfn"? Sure. > >> --- a/xen/common/page_alloc.c >> +++ b/xen/common/page_alloc.c >> @@ -2121,9 +2121,9 @@ void init_xenheap_pages(paddr_t ps, paddr_t pe) >> * Yuk! Ensure there is a one-page buffer between Xen and Dom zones, to >> * prevent merging of power-of-two blocks across the zone boundary. >> */ >> - if ( ps && !is_xen_heap_mfn(paddr_to_pfn(ps)-1) ) >> + if ( ps && !is_xen_heap_mfn(_mfn(paddr_to_pfn(ps)-1)) ) > > Please add the missing blanks around - at the same time. Ok. > >> ps += PAGE_SIZE; >> - if ( !is_xen_heap_mfn(paddr_to_pfn(pe)) ) >> + if ( !is_xen_heap_mfn(maddr_to_mfn(pe)) ) > > Why maddr_to_mfn() here but still paddr_to_pfn() above? Oh, > we don't have any mfn_sub(), I see. Yes we don't have mfn_sub() (or even gfn_sub()). I only found a couple of places where such helpers might be useful. I can introduce the 2 helpers if you think it is worth it. > >> --- a/xen/include/asm-x86/mm.h >> +++ b/xen/include/asm-x86/mm.h >> @@ -278,7 +278,7 @@ struct page_info >> >> #define is_xen_heap_page(page) ((page)->count_info & PGC_xen_heap) >> #define is_xen_heap_mfn(mfn) \ >> - (__mfn_valid(mfn) && is_xen_heap_page(mfn_to_page(_mfn(mfn)))) >> + (__mfn_valid(mfn_x(mfn)) && is_xen_heap_page(mfn_to_page(mfn))) > > Please don't open code mfn_valid(). That's a mistake. Someone I thought there were place in the code that was overriding mfn_valid(). I will update it. > > With these minor issues taken care of > Reviewed-by: Jan Beulich <jbeulich@suse.com> Thank you! > > Jan > >
>>> On 13.03.19 at 18:24, <julien.grall@arm.com> wrote: > On 13/03/2019 15:04, Jan Beulich wrote: >>>>> On 18.02.19 at 12:35, <julien.grall@arm.com> wrote: >>> --- a/xen/common/page_alloc.c >>> +++ b/xen/common/page_alloc.c >>> @@ -2121,9 +2121,9 @@ void init_xenheap_pages(paddr_t ps, paddr_t pe) >>> * Yuk! Ensure there is a one-page buffer between Xen and Dom zones, to >>> * prevent merging of power-of-two blocks across the zone boundary. >>> */ >>> - if ( ps && !is_xen_heap_mfn(paddr_to_pfn(ps)-1) ) >>> + if ( ps && !is_xen_heap_mfn(_mfn(paddr_to_pfn(ps)-1)) ) >>> ps += PAGE_SIZE; >>> - if ( !is_xen_heap_mfn(paddr_to_pfn(pe)) ) >>> + if ( !is_xen_heap_mfn(maddr_to_mfn(pe)) ) >> >> Why maddr_to_mfn() here but still paddr_to_pfn() above? Oh, >> we don't have any mfn_sub(), I see. > > Yes we don't have mfn_sub() (or even gfn_sub()). I only found a couple of places > where such helpers might be useful. I can introduce the 2 helpers if you think > it is worth it. Well, I guess in the end I'm fine either way. It simply struck me as odd at the first glance that you use maddr_to_mfn() in one case but not the other. Jan
Hi Jan, On 3/14/19 7:52 AM, Jan Beulich wrote: >>>> On 13.03.19 at 18:24, <julien.grall@arm.com> wrote: >> On 13/03/2019 15:04, Jan Beulich wrote: >>>>>> On 18.02.19 at 12:35, <julien.grall@arm.com> wrote: >>>> --- a/xen/common/page_alloc.c >>>> +++ b/xen/common/page_alloc.c >>>> @@ -2121,9 +2121,9 @@ void init_xenheap_pages(paddr_t ps, paddr_t pe) >>>> * Yuk! Ensure there is a one-page buffer between Xen and Dom zones, to >>>> * prevent merging of power-of-two blocks across the zone boundary. >>>> */ >>>> - if ( ps && !is_xen_heap_mfn(paddr_to_pfn(ps)-1) ) >>>> + if ( ps && !is_xen_heap_mfn(_mfn(paddr_to_pfn(ps)-1)) ) >>>> ps += PAGE_SIZE; >>>> - if ( !is_xen_heap_mfn(paddr_to_pfn(pe)) ) >>>> + if ( !is_xen_heap_mfn(maddr_to_mfn(pe)) ) >>> >>> Why maddr_to_mfn() here but still paddr_to_pfn() above? Oh, >>> we don't have any mfn_sub(), I see. >> >> Yes we don't have mfn_sub() (or even gfn_sub()). I only found a couple of places >> where such helpers might be useful. I can introduce the 2 helpers if you think >> it is worth it. > > Well, I guess in the end I'm fine either way. It simply struck me > as odd at the first glance that you use maddr_to_mfn() in one > case but not the other. I wanted to avoid adding mfn_x(...) on the line: _mfn(mfn_x(maddr_to_mfn(ps)) - 1) I will look to introduce mfn_sub()/gfn_sub(). Cheers,
On 14/03/2019 10:12, Julien Grall wrote: > Hi Jan, > > On 3/14/19 7:52 AM, Jan Beulich wrote: >>>>> On 13.03.19 at 18:24, <julien.grall@arm.com> wrote: >>> On 13/03/2019 15:04, Jan Beulich wrote: >>>>>>> On 18.02.19 at 12:35, <julien.grall@arm.com> wrote: >>>>> --- a/xen/common/page_alloc.c >>>>> +++ b/xen/common/page_alloc.c >>>>> @@ -2121,9 +2121,9 @@ void init_xenheap_pages(paddr_t ps, paddr_t pe) >>>>> * Yuk! Ensure there is a one-page buffer between Xen and >>>>> Dom zones, to >>>>> * prevent merging of power-of-two blocks across the zone >>>>> boundary. >>>>> */ >>>>> - if ( ps && !is_xen_heap_mfn(paddr_to_pfn(ps)-1) ) >>>>> + if ( ps && !is_xen_heap_mfn(_mfn(paddr_to_pfn(ps)-1)) ) >>>>> ps += PAGE_SIZE; >>>>> - if ( !is_xen_heap_mfn(paddr_to_pfn(pe)) ) >>>>> + if ( !is_xen_heap_mfn(maddr_to_mfn(pe)) ) >>>> >>>> Why maddr_to_mfn() here but still paddr_to_pfn() above? Oh, >>>> we don't have any mfn_sub(), I see. >>> >>> Yes we don't have mfn_sub() (or even gfn_sub()). I only found a >>> couple of places >>> where such helpers might be useful. I can introduce the 2 helpers if >>> you think >>> it is worth it. >> >> Well, I guess in the end I'm fine either way. It simply struck me >> as odd at the first glance that you use maddr_to_mfn() in one >> case but not the other. > > I wanted to avoid adding mfn_x(...) on the line: > > _mfn(mfn_x(maddr_to_mfn(ps)) - 1) > > I will look to introduce mfn_sub()/gfn_sub(). You do know that mfn_add(mfn, -1) will DTRT? ~Andrew
Hi Andrew, On 3/14/19 10:14 AM, Andrew Cooper wrote: > On 14/03/2019 10:12, Julien Grall wrote: >> Hi Jan, >> >> On 3/14/19 7:52 AM, Jan Beulich wrote: >>>>>> On 13.03.19 at 18:24, <julien.grall@arm.com> wrote: >>>> On 13/03/2019 15:04, Jan Beulich wrote: >>>>>>>> On 18.02.19 at 12:35, <julien.grall@arm.com> wrote: >>>>>> --- a/xen/common/page_alloc.c >>>>>> +++ b/xen/common/page_alloc.c >>>>>> @@ -2121,9 +2121,9 @@ void init_xenheap_pages(paddr_t ps, paddr_t pe) >>>>>> * Yuk! Ensure there is a one-page buffer between Xen and >>>>>> Dom zones, to >>>>>> * prevent merging of power-of-two blocks across the zone >>>>>> boundary. >>>>>> */ >>>>>> - if ( ps && !is_xen_heap_mfn(paddr_to_pfn(ps)-1) ) >>>>>> + if ( ps && !is_xen_heap_mfn(_mfn(paddr_to_pfn(ps)-1)) ) >>>>>> ps += PAGE_SIZE; >>>>>> - if ( !is_xen_heap_mfn(paddr_to_pfn(pe)) ) >>>>>> + if ( !is_xen_heap_mfn(maddr_to_mfn(pe)) ) >>>>> >>>>> Why maddr_to_mfn() here but still paddr_to_pfn() above? Oh, >>>>> we don't have any mfn_sub(), I see. >>>> >>>> Yes we don't have mfn_sub() (or even gfn_sub()). I only found a >>>> couple of places >>>> where such helpers might be useful. I can introduce the 2 helpers if >>>> you think >>>> it is worth it. >>> >>> Well, I guess in the end I'm fine either way. It simply struck me >>> as odd at the first glance that you use maddr_to_mfn() in one >>> case but not the other. >> >> I wanted to avoid adding mfn_x(...) on the line: >> >> _mfn(mfn_x(maddr_to_mfn(ps)) - 1) >> >> I will look to introduce mfn_sub()/gfn_sub(). > > You do know that mfn_add(mfn, -1) will DTRT? It didn't occur to me until you said it. I will use that then. Cheers,
>>> On 14.03.19 at 11:14, <andrew.cooper3@citrix.com> wrote: > You do know that mfn_add(mfn, -1) will DTRT? It will, but imo it looks slightly odd in particular in e.g. for ( ...; ...; mfn_add(mfn, -1) ), hence I didn't suggest it. But I don't object either. Jan
On 14/03/2019 11:47, Jan Beulich wrote: >>>> On 14.03.19 at 11:14, <andrew.cooper3@citrix.com> wrote: >> You do know that mfn_add(mfn, -1) will DTRT? > It will, but imo it looks slightly odd in particular in e.g. > for ( ...; ...; mfn_add(mfn, -1) ), hence I didn't suggest it. But > I don't object either. Well - in mathematics, x - 1 is identical to x + (-1), so the construct doesn't look overly odd to me. Alternatively, we could rename mfn_add to something else, but nothing comes to mind which would be as clear. ~Andrew
On Mon, 18 Feb 2019, Julien Grall wrote: > diff --git a/xen/include/asm-arm/mm.h b/xen/include/asm-arm/mm.h > index b56018aace..a9c8352b94 100644 > --- a/xen/include/asm-arm/mm.h > +++ b/xen/include/asm-arm/mm.h > @@ -138,16 +138,16 @@ extern vaddr_t xenheap_virt_start; > #endif > > #ifdef CONFIG_ARM_32 > -#define is_xen_heap_page(page) is_xen_heap_mfn(mfn_x(page_to_mfn(page))) > +#define is_xen_heap_page(page) is_xen_heap_mfn(page_to_mfn(page)) > #define is_xen_heap_mfn(mfn) ({ \ > - unsigned long mfn_ = (mfn); \ > + unsigned long mfn_ = mfn_x(mfn); \ > (mfn_ >= mfn_x(xenheap_mfn_start) && \ > mfn_ < mfn_x(xenheap_mfn_end)); \ > }) > #else > #define is_xen_heap_page(page) ((page)->count_info & PGC_xen_heap) > #define is_xen_heap_mfn(mfn) \ > - (mfn_valid(_mfn(mfn)) && is_xen_heap_page(mfn_to_page(_mfn(mfn)))) > + (mfn_valid(mfn) && is_xen_heap_page(mfn_to_page(mfn))) > #endif > > #define is_xen_fixed_mfn(mfn) \ > @@ -246,7 +246,7 @@ static inline paddr_t __virt_to_maddr(vaddr_t va) > #ifdef CONFIG_ARM_32 > static inline void *maddr_to_virt(paddr_t ma) > { > - ASSERT(is_xen_heap_mfn(ma >> PAGE_SHIFT)); > + ASSERT(is_xen_heap_mfn(maddr_to_mfn(ma))); > ma -= mfn_to_maddr(xenheap_mfn_start); > return (void *)(unsigned long) ma + XENHEAP_VIRT_START; > } For the ARM parts: Acked-by: Stefano Stabellini <sstabellini@kernel.org>
diff --git a/xen/arch/x86/mm.c b/xen/arch/x86/mm.c index df6e5bdd31..a1cd2fb421 100644 --- a/xen/arch/x86/mm.c +++ b/xen/arch/x86/mm.c @@ -4300,7 +4300,8 @@ int xenmem_add_to_physmap_one( { struct page_info *page = NULL; unsigned long gfn = 0; /* gcc ... */ - unsigned long prev_mfn, old_gpfn; + mfn_t prev_mfn; + unsigned long old_gpfn; int rc = 0; mfn_t mfn = INVALID_MFN; p2m_type_t p2mt; @@ -4349,12 +4350,12 @@ int xenmem_add_to_physmap_one( } /* Remove previously mapped page if it was present. */ - prev_mfn = mfn_x(get_gfn(d, gfn_x(gpfn), &p2mt)); - if ( mfn_valid(_mfn(prev_mfn)) ) + prev_mfn = get_gfn(d, gfn_x(gpfn), &p2mt); + if ( mfn_valid(prev_mfn) ) { if ( is_xen_heap_mfn(prev_mfn) ) /* Xen heap frames are simply unhooked from this phys slot. */ - rc = guest_physmap_remove_page(d, gpfn, _mfn(prev_mfn), PAGE_ORDER_4K); + rc = guest_physmap_remove_page(d, gpfn, prev_mfn, PAGE_ORDER_4K); else /* Normal domain memory is freed, to avoid leaking memory. */ rc = guest_remove_page(d, gfn_x(gpfn)); diff --git a/xen/arch/x86/mm/p2m.c b/xen/arch/x86/mm/p2m.c index d14ce57dd5..40fed111b9 100644 --- a/xen/arch/x86/mm/p2m.c +++ b/xen/arch/x86/mm/p2m.c @@ -2910,7 +2910,7 @@ int p2m_add_foreign(struct domain *tdom, unsigned long fgfn, prev_mfn = get_gfn(tdom, gpfn, &p2mt_prev); if ( mfn_valid(prev_mfn) ) { - if ( is_xen_heap_mfn(mfn_x(prev_mfn)) ) + if ( is_xen_heap_mfn(prev_mfn) ) /* Xen heap frames are simply unhooked from this phys slot */ rc = guest_physmap_remove_page(tdom, _gfn(gpfn), prev_mfn, 0); else diff --git a/xen/arch/x86/mm/shadow/multi.c b/xen/arch/x86/mm/shadow/multi.c index 7e9cbc69be..2d25e356d3 100644 --- a/xen/arch/x86/mm/shadow/multi.c +++ b/xen/arch/x86/mm/shadow/multi.c @@ -602,7 +602,7 @@ _sh_propagate(struct vcpu *v, * caching attributes in the shadows to match what was asked for. */ if ( (level == 1) && is_hvm_domain(d) && - !is_xen_heap_mfn(mfn_x(target_mfn)) ) + !is_xen_heap_mfn(target_mfn) ) { int type; diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c index 5de3686d85..e43c98d5cb 100644 --- a/xen/common/page_alloc.c +++ b/xen/common/page_alloc.c @@ -2121,9 +2121,9 @@ void init_xenheap_pages(paddr_t ps, paddr_t pe) * Yuk! Ensure there is a one-page buffer between Xen and Dom zones, to * prevent merging of power-of-two blocks across the zone boundary. */ - if ( ps && !is_xen_heap_mfn(paddr_to_pfn(ps)-1) ) + if ( ps && !is_xen_heap_mfn(_mfn(paddr_to_pfn(ps)-1)) ) ps += PAGE_SIZE; - if ( !is_xen_heap_mfn(paddr_to_pfn(pe)) ) + if ( !is_xen_heap_mfn(maddr_to_mfn(pe)) ) pe -= PAGE_SIZE; memguard_guard_range(maddr_to_virt(ps), pe - ps); diff --git a/xen/include/asm-arm/mm.h b/xen/include/asm-arm/mm.h index b56018aace..a9c8352b94 100644 --- a/xen/include/asm-arm/mm.h +++ b/xen/include/asm-arm/mm.h @@ -138,16 +138,16 @@ extern vaddr_t xenheap_virt_start; #endif #ifdef CONFIG_ARM_32 -#define is_xen_heap_page(page) is_xen_heap_mfn(mfn_x(page_to_mfn(page))) +#define is_xen_heap_page(page) is_xen_heap_mfn(page_to_mfn(page)) #define is_xen_heap_mfn(mfn) ({ \ - unsigned long mfn_ = (mfn); \ + unsigned long mfn_ = mfn_x(mfn); \ (mfn_ >= mfn_x(xenheap_mfn_start) && \ mfn_ < mfn_x(xenheap_mfn_end)); \ }) #else #define is_xen_heap_page(page) ((page)->count_info & PGC_xen_heap) #define is_xen_heap_mfn(mfn) \ - (mfn_valid(_mfn(mfn)) && is_xen_heap_page(mfn_to_page(_mfn(mfn)))) + (mfn_valid(mfn) && is_xen_heap_page(mfn_to_page(mfn))) #endif #define is_xen_fixed_mfn(mfn) \ @@ -246,7 +246,7 @@ static inline paddr_t __virt_to_maddr(vaddr_t va) #ifdef CONFIG_ARM_32 static inline void *maddr_to_virt(paddr_t ma) { - ASSERT(is_xen_heap_mfn(ma >> PAGE_SHIFT)); + ASSERT(is_xen_heap_mfn(maddr_to_mfn(ma))); ma -= mfn_to_maddr(xenheap_mfn_start); return (void *)(unsigned long) ma + XENHEAP_VIRT_START; } diff --git a/xen/include/asm-x86/mm.h b/xen/include/asm-x86/mm.h index f124f57964..1ca4154382 100644 --- a/xen/include/asm-x86/mm.h +++ b/xen/include/asm-x86/mm.h @@ -278,7 +278,7 @@ struct page_info #define is_xen_heap_page(page) ((page)->count_info & PGC_xen_heap) #define is_xen_heap_mfn(mfn) \ - (__mfn_valid(mfn) && is_xen_heap_page(mfn_to_page(_mfn(mfn)))) + (__mfn_valid(mfn_x(mfn)) && is_xen_heap_page(mfn_to_page(mfn))) #define is_xen_fixed_mfn(mfn) \ (((mfn_to_maddr(mfn)) >= __pa(&_stext)) && \ ((mfn_to_maddr(mfn)) <= __pa(&__2M_rwdata_end)))
No functional changes. Signed-off-by: Julien Grall <julien.grall@arm.com> --- xen/arch/x86/mm.c | 9 +++++---- xen/arch/x86/mm/p2m.c | 2 +- xen/arch/x86/mm/shadow/multi.c | 2 +- xen/common/page_alloc.c | 4 ++-- xen/include/asm-arm/mm.h | 8 ++++---- xen/include/asm-x86/mm.h | 2 +- 6 files changed, 14 insertions(+), 13 deletions(-)