| Message ID | 20190610221621.10938-3-hch@lst.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [01/17] mm: provide a print_vma_addr stub for !CONFIG_MMU | expand |
On 6/10/19 11:16 PM, Christoph Hellwig wrote: > The whole header file deals with swap entries and PTEs, none of which > can exist for nommu builds. Although I agree with the patch, I'm wondering how you get into it? Cheers Vladimir > > Signed-off-by: Christoph Hellwig <hch@lst.de> > --- > include/linux/swapops.h | 5 +++-- > 1 file changed, 3 insertions(+), 2 deletions(-) > > diff --git a/include/linux/swapops.h b/include/linux/swapops.h > index 4d961668e5fc..b02922556846 100644 > --- a/include/linux/swapops.h > +++ b/include/linux/swapops.h > @@ -6,6 +6,8 @@ > #include <linux/bug.h> > #include <linux/mm_types.h> > > +#ifdef CONFIG_MMU > + > /* > * swapcache pages are stored in the swapper_space radix tree. We want to > * get good packing density in that tree, so the index should be dense in > @@ -50,13 +52,11 @@ static inline pgoff_t swp_offset(swp_entry_t entry) > return entry.val & SWP_OFFSET_MASK; > } > > -#ifdef CONFIG_MMU > /* check whether a pte points to a swap entry */ > static inline int is_swap_pte(pte_t pte) > { > return !pte_none(pte) && !pte_present(pte); > } > -#endif > > /* > * Convert the arch-dependent pte representation of a swp_entry_t into an > @@ -375,4 +375,5 @@ static inline int non_swap_entry(swp_entry_t entry) > } > #endif > > +#endif /* CONFIG_MMU */ > #endif /* _LINUX_SWAPOPS_H */ >
On Tue, Jun 11, 2019 at 11:15:44AM +0100, Vladimir Murzin wrote: > On 6/10/19 11:16 PM, Christoph Hellwig wrote: > > The whole header file deals with swap entries and PTEs, none of which > > can exist for nommu builds. > > Although I agree with the patch, I'm wondering how you get into it? Without that the RISC-V nommu blows up like this: In file included from mm/vmscan.c:58: ./include/linux/swapops.h: In function ‘pte_to_swp_entry’: ./include/linux/swapops.h:71:15: error: implicit declaration of function ‘__pte_to_swp_entry’; did you mean ‘pte_to_swp_entry’? [-Werror=implicit-function-declaration] arch_entry = __pte_to_swp_entry(pte); ^~~~~~~~~~~~~~~~~~ pte_to_swp_entry ./include/linux/swapops.h:71:13: error: incompatible types when assigning to type ‘swp_entry_t’ {aka ‘struct <anonymous>’} from type ‘int’ arch_entry = __pte_to_swp_entry(pte); ^ ./include/linux/swapops.h:72:19: error: implicit declaration of function ‘__swp_type’; did you mean ‘swp_type’? [-Werror=implicit-function-declaration] return swp_entry(__swp_type(arch_entry), __swp_offset(arch_entry)); ^~~~~~~~~~ swp_type ./include/linux/swapops.h:72:43: error: implicit declaration of function ‘__swp_offset’; did you mean ‘swp_offset’? [-Werror=implicit-function-declaration] return swp_entry(__swp_type(arch_entry), __swp_offset(arch_entry)); ^~~~~~~~~~~~ swp_offset ./include/linux/swapops.h: In function ‘swp_entry_to_pte’: ./include/linux/swapops.h:83:15: error: implicit declaration of function ‘__swp_entry’; did you mean ‘swp_entry’? [-Werror=implicit-function-declaration] arch_entry = __swp_entry(swp_type(entry), swp_offset(entry)); ^~~~~~~~~~~ swp_entry ./include/linux/swapops.h:83:13: error: incompatible types when assigning to type ‘swp_entry_t’ {aka ‘struct <anonymous>’} from type ‘int’ arch_entry = __swp_entry(swp_type(entry), swp_offset(entry)); ^ ./include/linux/swapops.h:84:9: error: implicit declaration of function ‘__swp_entry_to_pte’; did you mean ‘swp_entry_to_pte’? [-Werror=implicit-function-declaration] return __swp_entry_to_pte(arch_entry); ^~~~~~~~~~~~~~~~~~ swp_entry_to_pte ./include/linux/swapops.h:84:9: error: incompatible types when returning type ‘int’ but ‘pte_t’ {aka ‘struct <anonymous>’} was expected return __swp_entry_to_pte(arch_entry); ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cc1: some warnings being treated as errors make[1]: *** [scripts/Makefile.build:278: mm/vmscan.o] Error 1 make: *** [Makefile:1071: mm] Error 2 make: *** Waiting for unfinished jobs....
On 6/11/19 3:18 PM, Christoph Hellwig wrote: > On Tue, Jun 11, 2019 at 11:15:44AM +0100, Vladimir Murzin wrote: >> On 6/10/19 11:16 PM, Christoph Hellwig wrote: >>> The whole header file deals with swap entries and PTEs, none of which >>> can exist for nommu builds. >> >> Although I agree with the patch, I'm wondering how you get into it? > > Without that the RISC-V nommu blows up like this: > > > In file included from mm/vmscan.c:58: > ./include/linux/swapops.h: In function ‘pte_to_swp_entry’: > ./include/linux/swapops.h:71:15: error: implicit declaration of function ‘__pte_to_swp_entry’; did you mean ‘pte_to_swp_entry’? [-Werror=implicit-function-declaration] > arch_entry = __pte_to_swp_entry(pte); > ^~~~~~~~~~~~~~~~~~ > pte_to_swp_entry > ./include/linux/swapops.h:71:13: error: incompatible types when assigning to type ‘swp_entry_t’ {aka ‘struct <anonymous>’} from type ‘int’ > arch_entry = __pte_to_swp_entry(pte); > ^ > ./include/linux/swapops.h:72:19: error: implicit declaration of function ‘__swp_type’; did you mean ‘swp_type’? [-Werror=implicit-function-declaration] > return swp_entry(__swp_type(arch_entry), __swp_offset(arch_entry)); > ^~~~~~~~~~ > swp_type > ./include/linux/swapops.h:72:43: error: implicit declaration of function ‘__swp_offset’; did you mean ‘swp_offset’? [-Werror=implicit-function-declaration] > return swp_entry(__swp_type(arch_entry), __swp_offset(arch_entry)); > ^~~~~~~~~~~~ > swp_offset > ./include/linux/swapops.h: In function ‘swp_entry_to_pte’: > ./include/linux/swapops.h:83:15: error: implicit declaration of function ‘__swp_entry’; did you mean ‘swp_entry’? [-Werror=implicit-function-declaration] > arch_entry = __swp_entry(swp_type(entry), swp_offset(entry)); > ^~~~~~~~~~~ > swp_entry > ./include/linux/swapops.h:83:13: error: incompatible types when assigning to type ‘swp_entry_t’ {aka ‘struct <anonymous>’} from type ‘int’ > arch_entry = __swp_entry(swp_type(entry), swp_offset(entry)); > ^ > ./include/linux/swapops.h:84:9: error: implicit declaration of function ‘__swp_entry_to_pte’; did you mean ‘swp_entry_to_pte’? [-Werror=implicit-function-declaration] > return __swp_entry_to_pte(arch_entry); > ^~~~~~~~~~~~~~~~~~ > swp_entry_to_pte > ./include/linux/swapops.h:84:9: error: incompatible types when returning type ‘int’ but ‘pte_t’ {aka ‘struct <anonymous>’} was expected > return __swp_entry_to_pte(arch_entry); > ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > cc1: some warnings being treated as errors > make[1]: *** [scripts/Makefile.build:278: mm/vmscan.o] Error 1 > make: *** [Makefile:1071: mm] Error 2 > make: *** Waiting for unfinished jobs.... > It looks like NOMMU ports tend to define those. For ARM they are: #define __swp_type(x) (0) #define __swp_offset(x) (0) #define __swp_entry(typ,off) ((swp_entry_t) { ((typ) | ((off) << 7)) }) #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) }) #define __swp_entry_to_pte(x) ((pte_t) { (x).val }) Anyway, I have no strong opinion on which is better :) Cheers Vladimir
On Tue, Jun 11, 2019 at 03:36:53PM +0100, Vladimir Murzin wrote: > It looks like NOMMU ports tend to define those. For ARM they are: > > #define __swp_type(x) (0) > #define __swp_offset(x) (0) > #define __swp_entry(typ,off) ((swp_entry_t) { ((typ) | ((off) << 7)) }) > #define __pte_to_swp_entry(pte) ((swp_entry_t) { pte_val(pte) }) > #define __swp_entry_to_pte(x) ((pte_t) { (x).val }) > > Anyway, I have no strong opinion on which is better :) It just seems a lot easier to stub out swapops.h rather than providing stubs in each arch so that inlines which we are never going to use can build. I can look into dropping this from the other nommu ports for the next merge window, though.
diff --git a/include/linux/swapops.h b/include/linux/swapops.h index 4d961668e5fc..b02922556846 100644 --- a/include/linux/swapops.h +++ b/include/linux/swapops.h @@ -6,6 +6,8 @@ #include <linux/bug.h> #include <linux/mm_types.h> +#ifdef CONFIG_MMU + /* * swapcache pages are stored in the swapper_space radix tree. We want to * get good packing density in that tree, so the index should be dense in @@ -50,13 +52,11 @@ static inline pgoff_t swp_offset(swp_entry_t entry) return entry.val & SWP_OFFSET_MASK; } -#ifdef CONFIG_MMU /* check whether a pte points to a swap entry */ static inline int is_swap_pte(pte_t pte) { return !pte_none(pte) && !pte_present(pte); } -#endif /* * Convert the arch-dependent pte representation of a swp_entry_t into an @@ -375,4 +375,5 @@ static inline int non_swap_entry(swp_entry_t entry) } #endif +#endif /* CONFIG_MMU */ #endif /* _LINUX_SWAPOPS_H */
The whole header file deals with swap entries and PTEs, none of which can exist for nommu builds. Signed-off-by: Christoph Hellwig <hch@lst.de> --- include/linux/swapops.h | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-)