Message ID | 5603F355.6070409@freescale.com |
---|---|
State | Superseded |
Headers | show |
On Thu, 2015-09-24 at 15:57 +0300, Laurentiu Tudor wrote: > Book-E MMUv2 present in e6500 cores supports > powers of 2K page sizes while older MMUv1 cores > support only powers of 4K page sizes, or in other > words the LSB of TSIZE on MMUv1 is always 0. > Thus, on MMUv2 we must not strip the LSB. We can get better TLB utilization by not stripping it, but why "must not" which makes it sound like a bugfix rather than an optimization? > Signed-off-by: Mihai Caraman <mihai.caraman@freescale.com> > [Laurentiu.Tudor@freescale.com: addressed review > feedback, split in distinct patch] > Signed-off-by: Laurentiu Tudor <Laurentiu.Tudor@freescale.com> > --- > arch/powerpc/kvm/e500_mmu_host.c | 28 +++++++++++++++++++++------- > 1 file changed, 21 insertions(+), 7 deletions(-) > > diff --git a/arch/powerpc/kvm/e500_mmu_host.c > b/arch/powerpc/kvm/e500_mmu_host.c > index 4d33e19..12d5c67 100644 > --- a/arch/powerpc/kvm/e500_mmu_host.c > +++ b/arch/powerpc/kvm/e500_mmu_host.c > @@ -371,6 +371,7 @@ static inline int kvmppc_e500_shadow_map(struct > kvmppc_vcpu_e500 *vcpu_e500, > > unsigned long start, end; > unsigned long slot_start, slot_end; > + int tsize_inc; > > pfnmap = 1; > > @@ -392,10 +393,20 @@ static inline int kvmppc_e500_shadow_map(struct > kvmppc_vcpu_e500 *vcpu_e500, > MAS1_TSIZE_SHIFT; > > /* > - * e500 doesn't implement the lowest tsize bit, > - * or 1K pages. > + * MMUv1 doesn't implement the lowest tsize bit, > + * or translations smaller than 4K. > */ > - tsize = max(BOOK3E_PAGESZ_4K, tsize & ~1); > + if (!has_feature(&vcpu_e500->vcpu, VCPU_FTR_MMU_V2)) > + tsize &= ~1; > + tsize = max(BOOK3E_PAGESZ_4K, tsize); > + > + /* > + * Calculate TSIZE increment. MMUv2 supports > + * power of 2K translations while MMUv1 is limited > + * to power of 4K sizes. > + */ > + tsize_inc = has_feature(&vcpu_e500->vcpu, > + VCPU_FTR_MMU_V2) ? 1 : 2; If you calculate tsize_inc first, then the previous if-statement can become "tsize &= ~(tsize_inc - 1);". > > /* > * Now find the largest tsize (up to what the guest > @@ -404,7 +415,8 @@ static inline int kvmppc_e500_shadow_map(struct > kvmppc_vcpu_e500 *vcpu_e500, > * aligned. > */ > > - for (; tsize > BOOK3E_PAGESZ_4K; tsize -= 2) { > + for (; tsize > BOOK3E_PAGESZ_4K; > + tsize -= tsize_inc) { > unsigned long gfn_start, gfn_end; > tsize_pages = 1 << (tsize - 2); > > @@ -437,10 +449,12 @@ static inline int kvmppc_e500_shadow_map(struct > kvmppc_vcpu_e500 *vcpu_e500, > tsize = min(__ilog2(psize) - 10, tsize); > > /* > - * e500 doesn't implement the lowest tsize bit, > - * or 1K pages. > + * MMUv1 doesn't implement the lowest tsize bit, > + * or translations smaller than 4K. > */ This comment makes it sound like MMUv2 might support translations smaller than 4K. -Scott -- To unsubscribe from this list: send the line "unsubscribe kvm-ppc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 09/24/2015 11:23 PM, Scott Wood wrote: > On Thu, 2015-09-24 at 15:57 +0300, Laurentiu Tudor wrote: >> Book-E MMUv2 present in e6500 cores supports >> powers of 2K page sizes while older MMUv1 cores >> support only powers of 4K page sizes, or in other >> words the LSB of TSIZE on MMUv1 is always 0. >> Thus, on MMUv2 we must not strip the LSB. > > We can get better TLB utilization by not stripping it, but why "must not" > which makes it sound like a bugfix rather than an optimization? Not sure i get it. If a guests wants a 2K^^(2n+1) translation size it will instead get a 2K^^2n size, no? Isn't this an issue? >> Signed-off-by: Mihai Caraman <mihai.caraman@freescale.com> >> [Laurentiu.Tudor@freescale.com: addressed review >> feedback, split in distinct patch] >> Signed-off-by: Laurentiu Tudor <Laurentiu.Tudor@freescale.com> >> --- >> arch/powerpc/kvm/e500_mmu_host.c | 28 +++++++++++++++++++++------- >> 1 file changed, 21 insertions(+), 7 deletions(-) >> >> diff --git a/arch/powerpc/kvm/e500_mmu_host.c >> b/arch/powerpc/kvm/e500_mmu_host.c >> index 4d33e19..12d5c67 100644 >> --- a/arch/powerpc/kvm/e500_mmu_host.c >> +++ b/arch/powerpc/kvm/e500_mmu_host.c >> @@ -371,6 +371,7 @@ static inline int kvmppc_e500_shadow_map(struct >> kvmppc_vcpu_e500 *vcpu_e500, >> >> unsigned long start, end; >> unsigned long slot_start, slot_end; >> + int tsize_inc; >> >> pfnmap = 1; >> >> @@ -392,10 +393,20 @@ static inline int kvmppc_e500_shadow_map(struct >> kvmppc_vcpu_e500 *vcpu_e500, >> MAS1_TSIZE_SHIFT; >> >> /* >> - * e500 doesn't implement the lowest tsize bit, >> - * or 1K pages. >> + * MMUv1 doesn't implement the lowest tsize bit, >> + * or translations smaller than 4K. >> */ >> - tsize = max(BOOK3E_PAGESZ_4K, tsize & ~1); >> + if (!has_feature(&vcpu_e500->vcpu, VCPU_FTR_MMU_V2)) >> + tsize &= ~1; >> + tsize = max(BOOK3E_PAGESZ_4K, tsize); >> + >> + /* >> + * Calculate TSIZE increment. MMUv2 supports >> + * power of 2K translations while MMUv1 is limited >> + * to power of 4K sizes. >> + */ >> + tsize_inc = has_feature(&vcpu_e500->vcpu, >> + VCPU_FTR_MMU_V2) ? 1 : 2; > > If you calculate tsize_inc first, then the previous if-statement can become > "tsize &= ~(tsize_inc - 1);". Thanks for the tip. Will do this in the next version. >> >> /* >> * Now find the largest tsize (up to what the guest >> @@ -404,7 +415,8 @@ static inline int kvmppc_e500_shadow_map(struct >> kvmppc_vcpu_e500 *vcpu_e500, >> * aligned. >> */ >> >> - for (; tsize > BOOK3E_PAGESZ_4K; tsize -= 2) { >> + for (; tsize > BOOK3E_PAGESZ_4K; >> + tsize -= tsize_inc) { >> unsigned long gfn_start, gfn_end; >> tsize_pages = 1 << (tsize - 2); >> >> @@ -437,10 +449,12 @@ static inline int kvmppc_e500_shadow_map(struct >> kvmppc_vcpu_e500 *vcpu_e500, >> tsize = min(__ilog2(psize) - 10, tsize); >> >> /* >> - * e500 doesn't implement the lowest tsize bit, >> - * or 1K pages. >> + * MMUv1 doesn't implement the lowest tsize bit, >> + * or translations smaller than 4K. >> */ > > This comment makes it sound like MMUv2 might support translations smaller > than 4K. Right. I'll rephrase in the next spin. Thanks for the review! --- Best Regards, Laurentiu -- To unsubscribe from this list: send the line "unsubscribe kvm-ppc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, 2015-09-25 at 17:30 +0300, Laurentiu Tudor wrote: > On 09/24/2015 11:23 PM, Scott Wood wrote: > > On Thu, 2015-09-24 at 15:57 +0300, Laurentiu Tudor wrote: > > > Book-E MMUv2 present in e6500 cores supports > > > powers of 2K page sizes while older MMUv1 cores > > > support only powers of 4K page sizes, or in other > > > words the LSB of TSIZE on MMUv1 is always 0. > > > Thus, on MMUv2 we must not strip the LSB. > > > > We can get better TLB utilization by not stripping it, but why "must not" > > which makes it sound like a bugfix rather than an optimization? > > Not sure i get it. If a guests wants a 2K^^(2n+1) translation > size it will instead get a 2K^^2n size, no? Isn't this an issue? It will get two pages of 2k^^2n size (assuming both halves are accessed). It will work, it will just consume twice as many TLB1 entries. It's the same as if the host pages backing the region are smaller than the mapping that the guest tries to create. -Scott -- To unsubscribe from this list: send the line "unsubscribe kvm-ppc" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/arch/powerpc/kvm/e500_mmu_host.c b/arch/powerpc/kvm/e500_mmu_host.c index 4d33e19..12d5c67 100644 --- a/arch/powerpc/kvm/e500_mmu_host.c +++ b/arch/powerpc/kvm/e500_mmu_host.c @@ -371,6 +371,7 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500, unsigned long start, end; unsigned long slot_start, slot_end; + int tsize_inc; pfnmap = 1; @@ -392,10 +393,20 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500, MAS1_TSIZE_SHIFT; /* - * e500 doesn't implement the lowest tsize bit, - * or 1K pages. + * MMUv1 doesn't implement the lowest tsize bit, + * or translations smaller than 4K. */ - tsize = max(BOOK3E_PAGESZ_4K, tsize & ~1); + if (!has_feature(&vcpu_e500->vcpu, VCPU_FTR_MMU_V2)) + tsize &= ~1; + tsize = max(BOOK3E_PAGESZ_4K, tsize); + + /* + * Calculate TSIZE increment. MMUv2 supports + * power of 2K translations while MMUv1 is limited + * to power of 4K sizes. + */ + tsize_inc = has_feature(&vcpu_e500->vcpu, + VCPU_FTR_MMU_V2) ? 1 : 2; /* * Now find the largest tsize (up to what the guest @@ -404,7 +415,8 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500, * aligned. */ - for (; tsize > BOOK3E_PAGESZ_4K; tsize -= 2) { + for (; tsize > BOOK3E_PAGESZ_4K; + tsize -= tsize_inc) { unsigned long gfn_start, gfn_end; tsize_pages = 1 << (tsize - 2); @@ -437,10 +449,12 @@ static inline int kvmppc_e500_shadow_map(struct kvmppc_vcpu_e500 *vcpu_e500, tsize = min(__ilog2(psize) - 10, tsize); /* - * e500 doesn't implement the lowest tsize bit, - * or 1K pages. + * MMUv1 doesn't implement the lowest tsize bit, + * or translations smaller than 4K. */ - tsize = max(BOOK3E_PAGESZ_4K, tsize & ~1); + if (!has_feature(&vcpu_e500->vcpu, VCPU_FTR_MMU_V2)) + tsize &= ~1; + tsize = max(BOOK3E_PAGESZ_4K, tsize); } up_read(¤t->mm->mmap_sem);