On Thu, Oct 17, 2024 at 10:54 AM Heiko Carstens <hca@xxxxxxxxxxxxx> wrote: > > On Thu, Oct 17, 2024 at 10:35:27AM -0700, Andrii Nakryiko wrote: > > On Thu, Oct 17, 2024 at 9:35 AM Shakeel Butt <shakeel.butt@xxxxxxxxx> wrote: > > > On Thu, Oct 17, 2024 at 11:18:34AM GMT, David Hildenbrand wrote: > > > > As replied elsewhere, can't we take a look at the mapping? > > > > > > > > We do the same thing in gup_fast_folio_allowed() where we check > > > > secretmem_mapping(). > > > > > > Responded on the v1 but I think we can go with v1 of this work as > > > whoever will be working on unmapping folios from direct map will need to > > > fix gup_fast_folio_allowed(), they can fix this code as well. Also it > > > seems like some arch don't have kernel_page_present() and builds are > > > failing. > > > > > > > Yeah, we are lucky that BPF CI tested s390x and caught this issue. > > > > > Andrii, let's move forward with the v1 patch. > > > > Let me post v3 based on v1 (checking for secretmem_mapping()), but > > I'll change return code to -EFAULT, so in the future this can be > > rolled into generic error handling code path with no change in error > > code. > > Ok, I've seen that you don't need kernel_page_present() anymore, just > after I implemented it for s390. I guess I'll send the patch below > (with a different commit message) upstream anyway, just in case > somebody else comes up with a similar use case. Please do send a patch, yes. It's good to have complete implementation of this API regardless. We can then switch to either kernel_page_present() or an alternative approach mentioned in [0] by David Hildenbrand, in the next release cycle, for instance. Thanks. [0] https://lore.kernel.org/all/c87a4ba0-b9c4-4044-b0c3-c1112601494f@xxxxxxxxxx/ > > From b625edc35de64293b728b030c62f7aaa65c8627e Mon Sep 17 00:00:00 2001 > From: Heiko Carstens <hca@xxxxxxxxxxxxx> > Date: Thu, 17 Oct 2024 19:41:07 +0200 > Subject: [PATCH] s390/pageattr: Implement missing kernel_page_present() > > kernel_page_present() was intentionally not implemented when adding > ARCH_HAS_SET_DIRECT_MAP support, since it was only used for suspend/resume > which is not supported anymore on s390. > > However a new bpf use case now leads to a compile error specific to > s390. Implement kernel_page_present() to fix this. > > Reported-by: Daniel Borkmann <daniel@xxxxxxxxxxxxx> > Closes: https://lore.kernel.org/all/045de961-ac69-40cc-b141-ab70ec9377ec@xxxxxxxxxxxxx > Fixes: 0490d6d7ba0a ("s390/mm: enable ARCH_HAS_SET_DIRECT_MAP") > Signed-off-by: Heiko Carstens <hca@xxxxxxxxxxxxx> > --- > arch/s390/include/asm/set_memory.h | 1 + > arch/s390/mm/pageattr.c | 15 +++++++++++++++ > 2 files changed, 16 insertions(+) > > diff --git a/arch/s390/include/asm/set_memory.h b/arch/s390/include/asm/set_memory.h > index 06fbabe2f66c..cb4cc0f59012 100644 > --- a/arch/s390/include/asm/set_memory.h > +++ b/arch/s390/include/asm/set_memory.h > @@ -62,5 +62,6 @@ __SET_MEMORY_FUNC(set_memory_4k, SET_MEMORY_4K) > > int set_direct_map_invalid_noflush(struct page *page); > int set_direct_map_default_noflush(struct page *page); > +bool kernel_page_present(struct page *page); > > #endif > diff --git a/arch/s390/mm/pageattr.c b/arch/s390/mm/pageattr.c > index 5f805ad42d4c..aec9eb16b6f7 100644 > --- a/arch/s390/mm/pageattr.c > +++ b/arch/s390/mm/pageattr.c > @@ -406,6 +406,21 @@ int set_direct_map_default_noflush(struct page *page) > return __set_memory((unsigned long)page_to_virt(page), 1, SET_MEMORY_DEF); > } > > +bool kernel_page_present(struct page *page) > +{ > + unsigned long addr; > + unsigned int cc; > + > + addr = (unsigned long)page_address(page); > + asm volatile( > + " lra %[addr],0(%[addr])\n" > + " ipm %[cc]\n" > + : [cc] "=d" (cc), [addr] "+a" (addr) > + : > + : "cc"); > + return (cc >> 28) == 0; > +} > + > #if defined(CONFIG_DEBUG_PAGEALLOC) || defined(CONFIG_KFENCE) > > static void ipte_range(pte_t *pte, unsigned long address, int nr) > -- > 2.45.2 >
