| Message ID | 20250316192445.2376484-4-luca.fancellu@arm.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | MPU mm subsystem skeleton | expand |
On 16/03/2025 20:24, Luca Fancellu wrote: > > > Introduce frame_table in order to provide the implementation of > virt_to_page for MPU system, move the MMU variant in mmu/mm.h. > > Introduce FRAMETABLE_NR that is required for 'pdx_group_valid' in > pdx.c, but leave the initialisation of the frame table to a later > stage. > Define FRAMETABLE_SIZE for MPU to support up to 1TB of ram, as the > only current implementation of armv8-r aarch64, which is cortex R82, > can address up to that memory. When mentioning support statements like this one, it'd be beneficial to provide a reference to a doc of some sort. Also, shouldn't this be occasion to clarify SUPPORT statement as for max RAM supported for ARMv8R-AArch64? ARMv8R support is experimental, so I'm not 100% sure if we need to provide support statement for it at this stage though. Better check with others. > > Take the occasion to sort alphabetically the headers following > the Xen code style and add the emacs footer in mpu/mm.c. > > Signed-off-by: Luca Fancellu <luca.fancellu@arm.com> > --- > v2 changes: > - sorted headers in mm.c > - modified commit message > - moved virt_to_page to MMU and MPU > - removed frametable_pdx_end, used mfn_valid > --- > xen/arch/arm/include/asm/mm.h | 14 -------------- > xen/arch/arm/include/asm/mmu/mm.h | 14 ++++++++++++++ > xen/arch/arm/include/asm/mpu/layout.h | 3 +++ > xen/arch/arm/include/asm/mpu/mm.h | 15 +++++++++++++++ > xen/arch/arm/mpu/mm.c | 14 +++++++++++++- > 5 files changed, 45 insertions(+), 15 deletions(-) > > diff --git a/xen/arch/arm/include/asm/mm.h b/xen/arch/arm/include/asm/mm.h > index 444fd03823ec..fbffaccef49b 100644 > --- a/xen/arch/arm/include/asm/mm.h > +++ b/xen/arch/arm/include/asm/mm.h > @@ -294,20 +294,6 @@ static inline uint64_t gvirt_to_maddr(vaddr_t va, paddr_t *pa, > #error "Unknown memory management layout" > #endif > > -/* Convert between Xen-heap virtual addresses and page-info structures. */ > -static inline struct page_info *virt_to_page(const void *v) > -{ > - unsigned long va = (unsigned long)v; > - unsigned long pdx; > - > - ASSERT(va >= XENHEAP_VIRT_START); > - ASSERT(va < directmap_virt_end); > - > - pdx = (va - XENHEAP_VIRT_START) >> PAGE_SHIFT; > - pdx += mfn_to_pdx(directmap_mfn_start); > - return frame_table + pdx - frametable_base_pdx; > -} > - > static inline void *page_to_virt(const struct page_info *pg) > { > return mfn_to_virt(mfn_x(page_to_mfn(pg))); > diff --git a/xen/arch/arm/include/asm/mmu/mm.h b/xen/arch/arm/include/asm/mmu/mm.h > index 6737c3ede783..caba987edc85 100644 > --- a/xen/arch/arm/include/asm/mmu/mm.h > +++ b/xen/arch/arm/include/asm/mmu/mm.h > @@ -70,6 +70,20 @@ static inline void *maddr_to_virt(paddr_t ma) > } > #endif > > +/* Convert between Xen-heap virtual addresses and page-info structures. */ > +static inline struct page_info *virt_to_page(const void *v) > +{ > + unsigned long va = (unsigned long)v; > + unsigned long pdx; > + > + ASSERT(va >= XENHEAP_VIRT_START); > + ASSERT(va < directmap_virt_end); > + > + pdx = (va - XENHEAP_VIRT_START) >> PAGE_SHIFT; > + pdx += mfn_to_pdx(directmap_mfn_start); > + return frame_table + pdx - frametable_base_pdx; > +} > + > /* > * Print a walk of a page table or p2m > * > diff --git a/xen/arch/arm/include/asm/mpu/layout.h b/xen/arch/arm/include/asm/mpu/layout.h > index 248e55f8882d..c331d1feaa84 100644 > --- a/xen/arch/arm/include/asm/mpu/layout.h > +++ b/xen/arch/arm/include/asm/mpu/layout.h > @@ -3,6 +3,9 @@ > #ifndef __ARM_MPU_LAYOUT_H__ > #define __ARM_MPU_LAYOUT_H__ > > +#define FRAMETABLE_SIZE GB(16) > +#define FRAMETABLE_NR (FRAMETABLE_SIZE / sizeof(*frame_table)) > + > #define XEN_START_ADDRESS CONFIG_XEN_START_ADDRESS > > /* > diff --git a/xen/arch/arm/include/asm/mpu/mm.h b/xen/arch/arm/include/asm/mpu/mm.h > index 6cfd0f5cd2c2..3a0a60dbfa18 100644 > --- a/xen/arch/arm/include/asm/mpu/mm.h > +++ b/xen/arch/arm/include/asm/mpu/mm.h > @@ -3,9 +3,13 @@ > #ifndef __ARM_MPU_MM_H__ > #define __ARM_MPU_MM_H__ > > +#include <xen/bug.h> > #include <xen/macros.h> > #include <xen/page-size.h> > #include <xen/types.h> > +#include <asm/mm.h> > + > +extern struct page_info *frame_table; > > #define virt_to_maddr(va) ((paddr_t)((vaddr_t)(va) & PADDR_MASK)) > > @@ -15,6 +19,17 @@ static inline void *maddr_to_virt(paddr_t ma) > return _p(ma); > } > > +/* Convert between virtual address to page-info structure. */ > +static inline struct page_info *virt_to_page(const void *v) > +{ > + paddr_t paddr = virt_to_maddr(v); > + unsigned long pdx = paddr_to_pdx(paddr); > + > + ASSERT(mfn_valid(maddr_to_mfn(paddr))); > + > + return frame_table + pdx - frametable_base_pdx; > +} This could be simplified (and number of conversions reduced) by doing sth like: mfn_t mfn = virt_to_mfn(v); ASSERT(mfn_valid(mfn)); return mfn_to_page(mfn); Other than that: Reviewed-by: Michal Orzel <michal.orzel@amd.com> ~Michal
Hi Michal, > On 17 Mar 2025, at 09:29, Orzel, Michal <michal.orzel@amd.com> wrote: > > > > On 16/03/2025 20:24, Luca Fancellu wrote: >> >> >> Introduce frame_table in order to provide the implementation of >> virt_to_page for MPU system, move the MMU variant in mmu/mm.h. >> >> Introduce FRAMETABLE_NR that is required for 'pdx_group_valid' in >> pdx.c, but leave the initialisation of the frame table to a later >> stage. >> Define FRAMETABLE_SIZE for MPU to support up to 1TB of ram, as the >> only current implementation of armv8-r aarch64, which is cortex R82, >> can address up to that memory. > When mentioning support statements like this one, it'd be beneficial to provide > a reference to a doc of some sort. So the only reference I have is this: https://developer.arm.com/Processors/Cortex-R82 but I would not be confident to use the link in the commit message as it could go stale very quickly. So I’m not sure about what I can do more. > > Also, shouldn't this be occasion to clarify SUPPORT statement as for max RAM > supported for ARMv8R-AArch64? ARMv8R support is experimental, so I'm not 100% > sure if we need to provide support statement for it at this stage though. Better > check with others. > Ok, I’ll stay tuned for the opinion of the others >> >> diff --git a/xen/arch/arm/include/asm/mpu/mm.h b/xen/arch/arm/include/asm/mpu/mm.h >> index 6cfd0f5cd2c2..3a0a60dbfa18 100644 >> --- a/xen/arch/arm/include/asm/mpu/mm.h >> +++ b/xen/arch/arm/include/asm/mpu/mm.h >> @@ -3,9 +3,13 @@ >> #ifndef __ARM_MPU_MM_H__ >> #define __ARM_MPU_MM_H__ >> >> +#include <xen/bug.h> >> #include <xen/macros.h> >> #include <xen/page-size.h> >> #include <xen/types.h> >> +#include <asm/mm.h> >> + >> +extern struct page_info *frame_table; >> >> #define virt_to_maddr(va) ((paddr_t)((vaddr_t)(va) & PADDR_MASK)) >> >> @@ -15,6 +19,17 @@ static inline void *maddr_to_virt(paddr_t ma) >> return _p(ma); >> } >> >> +/* Convert between virtual address to page-info structure. */ >> +static inline struct page_info *virt_to_page(const void *v) >> +{ >> + paddr_t paddr = virt_to_maddr(v); >> + unsigned long pdx = paddr_to_pdx(paddr); >> + >> + ASSERT(mfn_valid(maddr_to_mfn(paddr))); >> + >> + return frame_table + pdx - frametable_base_pdx; >> +} > This could be simplified (and number of conversions reduced) by doing sth like: > mfn_t mfn = virt_to_mfn(v); > > ASSERT(mfn_valid(mfn)); > > return mfn_to_page(mfn); Right, I’ll test these and I’ll use them. > > Other than that: > Reviewed-by: Michal Orzel <michal.orzel@amd.com> Thanks! > > ~Michal >
On 17/03/2025 10:33, Luca Fancellu wrote: > > > Hi Michal, > >> On 17 Mar 2025, at 09:29, Orzel, Michal <michal.orzel@amd.com> wrote: >> >> >> >> On 16/03/2025 20:24, Luca Fancellu wrote: >>> >>> >>> Introduce frame_table in order to provide the implementation of >>> virt_to_page for MPU system, move the MMU variant in mmu/mm.h. >>> >>> Introduce FRAMETABLE_NR that is required for 'pdx_group_valid' in >>> pdx.c, but leave the initialisation of the frame table to a later >>> stage. >>> Define FRAMETABLE_SIZE for MPU to support up to 1TB of ram, as the >>> only current implementation of armv8-r aarch64, which is cortex R82, >>> can address up to that memory. >> When mentioning support statements like this one, it'd be beneficial to provide >> a reference to a doc of some sort. > > So the only reference I have is this: https://developer.arm.com/Processors/Cortex-R82 > > but I would not be confident to use the link in the commit message as it could go stale > very quickly. So I’m not sure about what I can do more. Well, not really. Max physical memory is advertised via ID_AA64MMFR0_EL1. I found some old R82 technical manual (you can surely find the latest one and provide reference to it - not the web page) and indeed it mentions PARange as 0b0010 which is 40bit which is 1TB. With the R82 being the only CPU model implementing ARMv8R-AArch64, that's solid information. ~Michal
> On 17 Mar 2025, at 09:51, Orzel, Michal <michal.orzel@amd.com> wrote: > > > > On 17/03/2025 10:33, Luca Fancellu wrote: >> >> >> Hi Michal, >> >>> On 17 Mar 2025, at 09:29, Orzel, Michal <michal.orzel@amd.com> wrote: >>> >>> >>> >>> On 16/03/2025 20:24, Luca Fancellu wrote: >>>> >>>> >>>> Introduce frame_table in order to provide the implementation of >>>> virt_to_page for MPU system, move the MMU variant in mmu/mm.h. >>>> >>>> Introduce FRAMETABLE_NR that is required for 'pdx_group_valid' in >>>> pdx.c, but leave the initialisation of the frame table to a later >>>> stage. >>>> Define FRAMETABLE_SIZE for MPU to support up to 1TB of ram, as the >>>> only current implementation of armv8-r aarch64, which is cortex R82, >>>> can address up to that memory. >>> When mentioning support statements like this one, it'd be beneficial to provide >>> a reference to a doc of some sort. >> >> So the only reference I have is this: https://developer.arm.com/Processors/Cortex-R82 >> >> but I would not be confident to use the link in the commit message as it could go stale >> very quickly. So I’m not sure about what I can do more. > Well, not really. Max physical memory is advertised via ID_AA64MMFR0_EL1. I > found some old R82 technical manual (you can surely find the latest one and > provide reference to it - not the web page) and indeed it mentions PARange as > 0b0010 which is 40bit which is 1TB. With the R82 being the only CPU model > implementing ARMv8R-AArch64, that's solid information. Right, I forgot about it, thanks for pointing that out, I’ll add a reference about it Cheers, Luca
diff --git a/xen/arch/arm/include/asm/mm.h b/xen/arch/arm/include/asm/mm.h index 444fd03823ec..fbffaccef49b 100644 --- a/xen/arch/arm/include/asm/mm.h +++ b/xen/arch/arm/include/asm/mm.h @@ -294,20 +294,6 @@ static inline uint64_t gvirt_to_maddr(vaddr_t va, paddr_t *pa, #error "Unknown memory management layout" #endif -/* Convert between Xen-heap virtual addresses and page-info structures. */ -static inline struct page_info *virt_to_page(const void *v) -{ - unsigned long va = (unsigned long)v; - unsigned long pdx; - - ASSERT(va >= XENHEAP_VIRT_START); - ASSERT(va < directmap_virt_end); - - pdx = (va - XENHEAP_VIRT_START) >> PAGE_SHIFT; - pdx += mfn_to_pdx(directmap_mfn_start); - return frame_table + pdx - frametable_base_pdx; -} - static inline void *page_to_virt(const struct page_info *pg) { return mfn_to_virt(mfn_x(page_to_mfn(pg))); diff --git a/xen/arch/arm/include/asm/mmu/mm.h b/xen/arch/arm/include/asm/mmu/mm.h index 6737c3ede783..caba987edc85 100644 --- a/xen/arch/arm/include/asm/mmu/mm.h +++ b/xen/arch/arm/include/asm/mmu/mm.h @@ -70,6 +70,20 @@ static inline void *maddr_to_virt(paddr_t ma) } #endif +/* Convert between Xen-heap virtual addresses and page-info structures. */ +static inline struct page_info *virt_to_page(const void *v) +{ + unsigned long va = (unsigned long)v; + unsigned long pdx; + + ASSERT(va >= XENHEAP_VIRT_START); + ASSERT(va < directmap_virt_end); + + pdx = (va - XENHEAP_VIRT_START) >> PAGE_SHIFT; + pdx += mfn_to_pdx(directmap_mfn_start); + return frame_table + pdx - frametable_base_pdx; +} + /* * Print a walk of a page table or p2m * diff --git a/xen/arch/arm/include/asm/mpu/layout.h b/xen/arch/arm/include/asm/mpu/layout.h index 248e55f8882d..c331d1feaa84 100644 --- a/xen/arch/arm/include/asm/mpu/layout.h +++ b/xen/arch/arm/include/asm/mpu/layout.h @@ -3,6 +3,9 @@ #ifndef __ARM_MPU_LAYOUT_H__ #define __ARM_MPU_LAYOUT_H__ +#define FRAMETABLE_SIZE GB(16) +#define FRAMETABLE_NR (FRAMETABLE_SIZE / sizeof(*frame_table)) + #define XEN_START_ADDRESS CONFIG_XEN_START_ADDRESS /* diff --git a/xen/arch/arm/include/asm/mpu/mm.h b/xen/arch/arm/include/asm/mpu/mm.h index 6cfd0f5cd2c2..3a0a60dbfa18 100644 --- a/xen/arch/arm/include/asm/mpu/mm.h +++ b/xen/arch/arm/include/asm/mpu/mm.h @@ -3,9 +3,13 @@ #ifndef __ARM_MPU_MM_H__ #define __ARM_MPU_MM_H__ +#include <xen/bug.h> #include <xen/macros.h> #include <xen/page-size.h> #include <xen/types.h> +#include <asm/mm.h> + +extern struct page_info *frame_table; #define virt_to_maddr(va) ((paddr_t)((vaddr_t)(va) & PADDR_MASK)) @@ -15,6 +19,17 @@ static inline void *maddr_to_virt(paddr_t ma) return _p(ma); } +/* Convert between virtual address to page-info structure. */ +static inline struct page_info *virt_to_page(const void *v) +{ + paddr_t paddr = virt_to_maddr(v); + unsigned long pdx = paddr_to_pdx(paddr); + + ASSERT(mfn_valid(maddr_to_mfn(paddr))); + + return frame_table + pdx - frametable_base_pdx; +} + #endif /* __ARM_MPU_MM_H__ */ /* diff --git a/xen/arch/arm/mpu/mm.c b/xen/arch/arm/mpu/mm.c index 0b8748e57598..3632011c1013 100644 --- a/xen/arch/arm/mpu/mm.c +++ b/xen/arch/arm/mpu/mm.c @@ -1,9 +1,12 @@ /* SPDX-License-Identifier: GPL-2.0-only */ -#include <xen/lib.h> #include <xen/init.h> +#include <xen/lib.h> +#include <xen/mm.h> #include <xen/sizes.h> +struct page_info *frame_table; + static void __init __maybe_unused build_assertions(void) { /* @@ -13,3 +16,12 @@ static void __init __maybe_unused build_assertions(void) */ BUILD_BUG_ON(PAGE_SIZE != SZ_4K); } + +/* + * Local variables: + * mode: C + * c-file-style: "BSD" + * c-basic-offset: 4 + * indent-tabs-mode: nil + * End: + */
Introduce frame_table in order to provide the implementation of virt_to_page for MPU system, move the MMU variant in mmu/mm.h. Introduce FRAMETABLE_NR that is required for 'pdx_group_valid' in pdx.c, but leave the initialisation of the frame table to a later stage. Define FRAMETABLE_SIZE for MPU to support up to 1TB of ram, as the only current implementation of armv8-r aarch64, which is cortex R82, can address up to that memory. Take the occasion to sort alphabetically the headers following the Xen code style and add the emacs footer in mpu/mm.c. Signed-off-by: Luca Fancellu <luca.fancellu@arm.com> --- v2 changes: - sorted headers in mm.c - modified commit message - moved virt_to_page to MMU and MPU - removed frametable_pdx_end, used mfn_valid --- xen/arch/arm/include/asm/mm.h | 14 -------------- xen/arch/arm/include/asm/mmu/mm.h | 14 ++++++++++++++ xen/arch/arm/include/asm/mpu/layout.h | 3 +++ xen/arch/arm/include/asm/mpu/mm.h | 15 +++++++++++++++ xen/arch/arm/mpu/mm.c | 14 +++++++++++++- 5 files changed, 45 insertions(+), 15 deletions(-)