| Message ID | 20240226120321.1055731-2-ryan.roberts@arm.com (mailing list archive) |
|---|---|
| State | New |
| Headers | show |
| Series | Address some contpte nits | expand |
On 26.02.24 13:03, Ryan Roberts wrote: > The contpte symbols must be exported since some of the public inline > ptep_* APIs are called from modules and these inlines now call the > contpte functions. Originally they were exported as EXPORT_SYMBOL() for > fear of breaking out-of-tree modules. But we subsequently concluded that > EXPORT_SYMBOL_GPL() should be safe since these functions are deeply core > mm routines, and any module operating at this level is not going to be > able to survive on EXPORT_SYMBOL alone. > I only looked at __set_ptes() to get a feeling what would currently work. __set_ptes() might already call __sync_icache_dcache() via __sync_cache_and_tags(), that is EXPORT_SYMBOL_GPL. [mte_sync_tags() is not exported at all, so maybe it's safe to assume that some out-of-tree module could not make good use of set_pte_at() in general] Acked-by: David Hildenbrand <david@redhat.com>
On 26/02/2024 12:25, David Hildenbrand wrote: > On 26.02.24 13:03, Ryan Roberts wrote: >> The contpte symbols must be exported since some of the public inline >> ptep_* APIs are called from modules and these inlines now call the >> contpte functions. Originally they were exported as EXPORT_SYMBOL() for >> fear of breaking out-of-tree modules. But we subsequently concluded that >> EXPORT_SYMBOL_GPL() should be safe since these functions are deeply core >> mm routines, and any module operating at this level is not going to be >> able to survive on EXPORT_SYMBOL alone. >> > > I only looked at __set_ptes() to get a feeling what would currently work. > > __set_ptes() might already call __sync_icache_dcache() via > __sync_cache_and_tags(), that is EXPORT_SYMBOL_GPL. > > [mte_sync_tags() is not exported at all, so maybe it's safe to assume that some > out-of-tree module could not make good use of set_pte_at() in general] That's interesting, some grepping I previously did showed that ptep_get() and set_pte_at() are used by in-tree drivers (which I assume can be built as modules): https://lore.kernel.org/linux-arm-kernel/b994ff89-1a1f-26ca-9479-b08c77f94be8@arm.com/ > > Acked-by: David Hildenbrand <david@redhat.com> > Thanks!
On 2/26/24 04:03, Ryan Roberts wrote: > The contpte symbols must be exported since some of the public inline > ptep_* APIs are called from modules and these inlines now call the > contpte functions. Originally they were exported as EXPORT_SYMBOL() for > fear of breaking out-of-tree modules. But we subsequently concluded that > EXPORT_SYMBOL_GPL() should be safe since these functions are deeply core > mm routines, and any module operating at this level is not going to be > able to survive on EXPORT_SYMBOL alone. > > Link: https://lore.kernel.org/linux-mm/f9fc2b31-11cb-4969-8961-9c89fea41b74@nvidia.com/ > Signed-off-by: Ryan Roberts <ryan.roberts@arm.com> > --- > arch/arm64/mm/contpte.c | 22 +++++++++++----------- > 1 file changed, 11 insertions(+), 11 deletions(-) > Yes, looks good. Reviewed-by: John Hubbard <jhubbard@nvidia.com> thanks,
On Mon, Feb 26, 2024 at 12:03:20PM +0000, Ryan Roberts wrote: > The contpte symbols must be exported since some of the public inline > ptep_* APIs are called from modules and these inlines now call the > contpte functions. Originally they were exported as EXPORT_SYMBOL() for > fear of breaking out-of-tree modules. But we subsequently concluded that > EXPORT_SYMBOL_GPL() should be safe since these functions are deeply core > mm routines, and any module operating at this level is not going to be > able to survive on EXPORT_SYMBOL alone. > > Link: https://lore.kernel.org/linux-mm/f9fc2b31-11cb-4969-8961-9c89fea41b74@nvidia.com/ > Signed-off-by: Ryan Roberts <ryan.roberts@arm.com> Acked-by: Catalin Marinas <catalin.marinas@arm.com>
diff --git a/arch/arm64/mm/contpte.c b/arch/arm64/mm/contpte.c index 16788f07716d..be0a226c4ff9 100644 --- a/arch/arm64/mm/contpte.c +++ b/arch/arm64/mm/contpte.c @@ -135,7 +135,7 @@ void __contpte_try_fold(struct mm_struct *mm, unsigned long addr, pte = pte_mkcont(pte); contpte_convert(mm, addr, orig_ptep, pte); } -EXPORT_SYMBOL(__contpte_try_fold); +EXPORT_SYMBOL_GPL(__contpte_try_fold); void __contpte_try_unfold(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte) @@ -150,7 +150,7 @@ void __contpte_try_unfold(struct mm_struct *mm, unsigned long addr, pte = pte_mknoncont(pte); contpte_convert(mm, addr, ptep, pte); } -EXPORT_SYMBOL(__contpte_try_unfold); +EXPORT_SYMBOL_GPL(__contpte_try_unfold); pte_t contpte_ptep_get(pte_t *ptep, pte_t orig_pte) { @@ -178,7 +178,7 @@ pte_t contpte_ptep_get(pte_t *ptep, pte_t orig_pte) return orig_pte; } -EXPORT_SYMBOL(contpte_ptep_get); +EXPORT_SYMBOL_GPL(contpte_ptep_get); pte_t contpte_ptep_get_lockless(pte_t *orig_ptep) { @@ -231,7 +231,7 @@ pte_t contpte_ptep_get_lockless(pte_t *orig_ptep) return orig_pte; } -EXPORT_SYMBOL(contpte_ptep_get_lockless); +EXPORT_SYMBOL_GPL(contpte_ptep_get_lockless); void contpte_set_ptes(struct mm_struct *mm, unsigned long addr, pte_t *ptep, pte_t pte, unsigned int nr) @@ -274,7 +274,7 @@ void contpte_set_ptes(struct mm_struct *mm, unsigned long addr, } while (addr != end); } -EXPORT_SYMBOL(contpte_set_ptes); +EXPORT_SYMBOL_GPL(contpte_set_ptes); void contpte_clear_full_ptes(struct mm_struct *mm, unsigned long addr, pte_t *ptep, unsigned int nr, int full) @@ -282,7 +282,7 @@ void contpte_clear_full_ptes(struct mm_struct *mm, unsigned long addr, contpte_try_unfold_partial(mm, addr, ptep, nr); __clear_full_ptes(mm, addr, ptep, nr, full); } -EXPORT_SYMBOL(contpte_clear_full_ptes); +EXPORT_SYMBOL_GPL(contpte_clear_full_ptes); pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm, unsigned long addr, pte_t *ptep, @@ -291,7 +291,7 @@ pte_t contpte_get_and_clear_full_ptes(struct mm_struct *mm, contpte_try_unfold_partial(mm, addr, ptep, nr); return __get_and_clear_full_ptes(mm, addr, ptep, nr, full); } -EXPORT_SYMBOL(contpte_get_and_clear_full_ptes); +EXPORT_SYMBOL_GPL(contpte_get_and_clear_full_ptes); int contpte_ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep) @@ -316,7 +316,7 @@ int contpte_ptep_test_and_clear_young(struct vm_area_struct *vma, return young; } -EXPORT_SYMBOL(contpte_ptep_test_and_clear_young); +EXPORT_SYMBOL_GPL(contpte_ptep_test_and_clear_young); int contpte_ptep_clear_flush_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep) @@ -337,7 +337,7 @@ int contpte_ptep_clear_flush_young(struct vm_area_struct *vma, return young; } -EXPORT_SYMBOL(contpte_ptep_clear_flush_young); +EXPORT_SYMBOL_GPL(contpte_ptep_clear_flush_young); void contpte_wrprotect_ptes(struct mm_struct *mm, unsigned long addr, pte_t *ptep, unsigned int nr) @@ -355,7 +355,7 @@ void contpte_wrprotect_ptes(struct mm_struct *mm, unsigned long addr, contpte_try_unfold_partial(mm, addr, ptep, nr); __wrprotect_ptes(mm, addr, ptep, nr); } -EXPORT_SYMBOL(contpte_wrprotect_ptes); +EXPORT_SYMBOL_GPL(contpte_wrprotect_ptes); int contpte_ptep_set_access_flags(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep, @@ -401,4 +401,4 @@ int contpte_ptep_set_access_flags(struct vm_area_struct *vma, return 1; } -EXPORT_SYMBOL(contpte_ptep_set_access_flags); +EXPORT_SYMBOL_GPL(contpte_ptep_set_access_flags);
The contpte symbols must be exported since some of the public inline ptep_* APIs are called from modules and these inlines now call the contpte functions. Originally they were exported as EXPORT_SYMBOL() for fear of breaking out-of-tree modules. But we subsequently concluded that EXPORT_SYMBOL_GPL() should be safe since these functions are deeply core mm routines, and any module operating at this level is not going to be able to survive on EXPORT_SYMBOL alone. Link: https://lore.kernel.org/linux-mm/f9fc2b31-11cb-4969-8961-9c89fea41b74@nvidia.com/ Signed-off-by: Ryan Roberts <ryan.roberts@arm.com> --- arch/arm64/mm/contpte.c | 22 +++++++++++----------- 1 file changed, 11 insertions(+), 11 deletions(-)