| Message ID | 1522636236-12625-2-git-send-email-hejianet@gmail.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 2 April 2018 at 04:30, Jia He <hejianet@gmail.com> wrote: > Commit b92df1de5d28 ("mm: page_alloc: skip over regions of invalid pfns > where possible") optimized the loop in memmap_init_zone(). But it causes > possible panic bug. So Daniel Vacek reverted it later. > > But as suggested by Daniel Vacek, it is fine to using memblock to skip > gaps and finding next valid frame with CONFIG_HAVE_ARCH_PFN_VALID. > > On arm and arm64, memblock is used by default. But generic version of > pfn_valid() is based on mem sections and memblock_next_valid_pfn() does > not always return the next valid one but skips more resulting in some > valid frames to be skipped (as if they were invalid). And that's why > kernel was eventually crashing on some !arm machines. > > And as verified by Eugeniu Rosca, arm can benifit from commit > b92df1de5d28. So remain the memblock_next_valid_pfn on arm{,64} and move > the related codes to arm64 arch directory. > > Suggested-by: Daniel Vacek <neelx@redhat.com> > Signed-off-by: Jia He <jia.he@hxt-semitech.com> Hello Jia, Apologies for chiming in late. If we are going to rearchitect this, I'd rather we change the loop in memmap_init_zone() so that we skip to the next valid PFN directly rather than skipping to the last invalid PFN so that the pfn++ in the for () results in the next value. Can we replace the pfn++ there with a function calls that defaults to 'return pfn + 1', but does the skip for architectures that implement it? > --- > arch/arm/include/asm/page.h | 2 ++ > arch/arm/mm/init.c | 31 ++++++++++++++++++++++++++++++- > arch/arm64/include/asm/page.h | 2 ++ > arch/arm64/mm/init.c | 31 ++++++++++++++++++++++++++++++- > include/linux/mmzone.h | 1 + > mm/page_alloc.c | 4 +++- > 6 files changed, 68 insertions(+), 3 deletions(-) > > diff --git a/arch/arm/include/asm/page.h b/arch/arm/include/asm/page.h > index 4355f0e..489875c 100644 > --- a/arch/arm/include/asm/page.h > +++ b/arch/arm/include/asm/page.h > @@ -158,6 +158,8 @@ typedef struct page *pgtable_t; > > #ifdef CONFIG_HAVE_ARCH_PFN_VALID > extern int pfn_valid(unsigned long); > +extern unsigned long memblock_next_valid_pfn(unsigned long pfn); > +#define skip_to_last_invalid_pfn(pfn) (memblock_next_valid_pfn(pfn) - 1) > #endif > > #include <asm/memory.h> > diff --git a/arch/arm/mm/init.c b/arch/arm/mm/init.c > index a1f11a7..0fb85ca 100644 > --- a/arch/arm/mm/init.c > +++ b/arch/arm/mm/init.c > @@ -198,7 +198,36 @@ int pfn_valid(unsigned long pfn) > return memblock_is_map_memory(__pfn_to_phys(pfn)); > } > EXPORT_SYMBOL(pfn_valid); > -#endif > + > +/* HAVE_MEMBLOCK is always enabled on arm */ > +unsigned long __init_memblock memblock_next_valid_pfn(unsigned long pfn) > +{ > + struct memblock_type *type = &memblock.memory; > + unsigned int right = type->cnt; > + unsigned int mid, left = 0; > + phys_addr_t addr = PFN_PHYS(++pfn); > + > + do { > + mid = (right + left) / 2; > + > + if (addr < type->regions[mid].base) > + right = mid; > + else if (addr >= (type->regions[mid].base + > + type->regions[mid].size)) > + left = mid + 1; > + else { > + /* addr is within the region, so pfn is valid */ > + return pfn; > + } > + } while (left < right); > + > + if (right == type->cnt) > + return -1UL; > + else > + return PHYS_PFN(type->regions[right].base); > +} > +EXPORT_SYMBOL(memblock_next_valid_pfn); > +#endif /*CONFIG_HAVE_ARCH_PFN_VALID*/ > > #ifndef CONFIG_SPARSEMEM > static void __init arm_memory_present(void) > diff --git a/arch/arm64/include/asm/page.h b/arch/arm64/include/asm/page.h > index 60d02c8..e57d3f2 100644 > --- a/arch/arm64/include/asm/page.h > +++ b/arch/arm64/include/asm/page.h > @@ -39,6 +39,8 @@ typedef struct page *pgtable_t; > > #ifdef CONFIG_HAVE_ARCH_PFN_VALID > extern int pfn_valid(unsigned long); > +extern unsigned long memblock_next_valid_pfn(unsigned long pfn); > +#define skip_to_last_invalid_pfn(pfn) (memblock_next_valid_pfn(pfn) - 1) > #endif > > #include <asm/memory.h> > diff --git a/arch/arm64/mm/init.c b/arch/arm64/mm/init.c > index 00e7b90..13e43ff 100644 > --- a/arch/arm64/mm/init.c > +++ b/arch/arm64/mm/init.c > @@ -290,7 +290,36 @@ int pfn_valid(unsigned long pfn) > return memblock_is_map_memory(pfn << PAGE_SHIFT); > } > EXPORT_SYMBOL(pfn_valid); > -#endif > + > +/* HAVE_MEMBLOCK is always enabled on arm64 */ > +unsigned long __init_memblock memblock_next_valid_pfn(unsigned long pfn) > +{ > + struct memblock_type *type = &memblock.memory; > + unsigned int right = type->cnt; > + unsigned int mid, left = 0; > + phys_addr_t addr = PFN_PHYS(++pfn); > + > + do { > + mid = (right + left) / 2; > + > + if (addr < type->regions[mid].base) > + right = mid; > + else if (addr >= (type->regions[mid].base + > + type->regions[mid].size)) > + left = mid + 1; > + else { > + /* addr is within the region, so pfn is valid */ > + return pfn; > + } > + } while (left < right); > + > + if (right == type->cnt) > + return -1UL; > + else > + return PHYS_PFN(type->regions[right].base); > +} > +EXPORT_SYMBOL(memblock_next_valid_pfn); > +#endif /*CONFIG_HAVE_ARCH_PFN_VALID*/ > > #ifndef CONFIG_SPARSEMEM > static void __init arm64_memory_present(void) > diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h > index d797716..f9c0c46 100644 > --- a/include/linux/mmzone.h > +++ b/include/linux/mmzone.h > @@ -1245,6 +1245,7 @@ static inline int pfn_valid(unsigned long pfn) > return 0; > return valid_section(__nr_to_section(pfn_to_section_nr(pfn))); > } > +#define skip_to_last_invalid_pfn(pfn) (pfn) > #endif > > static inline int pfn_present(unsigned long pfn) > diff --git a/mm/page_alloc.c b/mm/page_alloc.c > index c19f5ac..30f7d76 100644 > --- a/mm/page_alloc.c > +++ b/mm/page_alloc.c > @@ -5483,8 +5483,10 @@ void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone, > if (context != MEMMAP_EARLY) > goto not_early; > > - if (!early_pfn_valid(pfn)) > + if (!early_pfn_valid(pfn)) { > + pfn = skip_to_last_invalid_pfn(pfn); > continue; > + } > if (!early_pfn_in_nid(pfn, nid)) > continue; > if (!update_defer_init(pgdat, pfn, end_pfn, &nr_initialised)) > -- > 2.7.4 >
On 4/2/2018 2:55 PM, Ard Biesheuvel Wrote: > On 2 April 2018 at 04:30, Jia He <hejianet@gmail.com> wrote: >> Commit b92df1de5d28 ("mm: page_alloc: skip over regions of invalid pfns >> where possible") optimized the loop in memmap_init_zone(). But it causes >> possible panic bug. So Daniel Vacek reverted it later. >> >> But as suggested by Daniel Vacek, it is fine to using memblock to skip >> gaps and finding next valid frame with CONFIG_HAVE_ARCH_PFN_VALID. >> >> On arm and arm64, memblock is used by default. But generic version of >> pfn_valid() is based on mem sections and memblock_next_valid_pfn() does >> not always return the next valid one but skips more resulting in some >> valid frames to be skipped (as if they were invalid). And that's why >> kernel was eventually crashing on some !arm machines. >> >> And as verified by Eugeniu Rosca, arm can benifit from commit >> b92df1de5d28. So remain the memblock_next_valid_pfn on arm{,64} and move >> the related codes to arm64 arch directory. >> >> Suggested-by: Daniel Vacek <neelx@redhat.com> >> Signed-off-by: Jia He <jia.he@hxt-semitech.com> > Hello Jia, > > Apologies for chiming in late. no problem, thanks for your comments ;-) > > If we are going to rearchitect this, I'd rather we change the loop in > memmap_init_zone() so that we skip to the next valid PFN directly > rather than skipping to the last invalid PFN so that the pfn++ in the hmm... Maybe this macro name makes you confused pfn = skip_to_last_invalid_pfn(pfn); how about skip_to_next_valid_pfn? > for () results in the next value. Can we replace the pfn++ there with > a function calls that defaults to 'return pfn + 1', but does the skip > for architectures that implement it? I am not sure I understand your question here. With this patch, on !arm arches, skip_to_last_invalid_pfn is equal to (pfn), and will be increased when for{} loop continue. We only *skip* to the start pfn of next valid region when CONFIG_HAVE_MEMBLOCK and CONFIG_HAVE_ARCH_PFN_VALID(arm/arm64 supports both).
Hi Jia,
Thank you for the patch! Yet something to improve:
[auto build test ERROR on linus/master]
[also build test ERROR on v4.16 next-20180329]
[cannot apply to arm64/for-next/core]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]
url: https://github.com/0day-ci/linux/commits/Jia-He/optimize-memblock_next_valid_pfn-and-early_pfn_valid-on-arm-and-arm64/20180402-131223
config: i386-tinyconfig (attached as .config)
compiler: gcc-7 (Debian 7.3.0-1) 7.3.0
reproduce:
# save the attached .config to linux build tree
make ARCH=i386
All errors (new ones prefixed by >>):
mm/page_alloc.c: In function 'memmap_init_zone':
>> mm/page_alloc.c:5360:10: error: implicit declaration of function 'skip_to_last_invalid_pfn' [-Werror=implicit-function-declaration]
pfn = skip_to_last_invalid_pfn(pfn);
^~~~~~~~~~~~~~~~~~~~~~~~
cc1: some warnings being treated as errors
vim +/skip_to_last_invalid_pfn +5360 mm/page_alloc.c
5340
5341 if (highest_memmap_pfn < end_pfn - 1)
5342 highest_memmap_pfn = end_pfn - 1;
5343
5344 /*
5345 * Honor reservation requested by the driver for this ZONE_DEVICE
5346 * memory
5347 */
5348 if (altmap && start_pfn == altmap->base_pfn)
5349 start_pfn += altmap->reserve;
5350
5351 for (pfn = start_pfn; pfn < end_pfn; pfn++) {
5352 /*
5353 * There can be holes in boot-time mem_map[]s handed to this
5354 * function. They do not exist on hotplugged memory.
5355 */
5356 if (context != MEMMAP_EARLY)
5357 goto not_early;
5358
5359 if (!early_pfn_valid(pfn)) {
> 5360 pfn = skip_to_last_invalid_pfn(pfn);
5361 continue;
5362 }
5363 if (!early_pfn_in_nid(pfn, nid))
5364 continue;
5365 if (!update_defer_init(pgdat, pfn, end_pfn, &nr_initialised))
5366 break;
5367
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation
On 2 April 2018 at 09:49, Jia He <hejianet@gmail.com> wrote: > > > On 4/2/2018 2:55 PM, Ard Biesheuvel Wrote: >> >> On 2 April 2018 at 04:30, Jia He <hejianet@gmail.com> wrote: >>> >>> Commit b92df1de5d28 ("mm: page_alloc: skip over regions of invalid pfns >>> where possible") optimized the loop in memmap_init_zone(). But it causes >>> possible panic bug. So Daniel Vacek reverted it later. >>> >>> But as suggested by Daniel Vacek, it is fine to using memblock to skip >>> gaps and finding next valid frame with CONFIG_HAVE_ARCH_PFN_VALID. >>> >>> On arm and arm64, memblock is used by default. But generic version of >>> pfn_valid() is based on mem sections and memblock_next_valid_pfn() does >>> not always return the next valid one but skips more resulting in some >>> valid frames to be skipped (as if they were invalid). And that's why >>> kernel was eventually crashing on some !arm machines. >>> >>> And as verified by Eugeniu Rosca, arm can benifit from commit >>> b92df1de5d28. So remain the memblock_next_valid_pfn on arm{,64} and move >>> the related codes to arm64 arch directory. >>> >>> Suggested-by: Daniel Vacek <neelx@redhat.com> >>> Signed-off-by: Jia He <jia.he@hxt-semitech.com> >> >> Hello Jia, >> >> Apologies for chiming in late. > > no problem, thanks for your comments ;-) >> >> >> If we are going to rearchitect this, I'd rather we change the loop in >> memmap_init_zone() so that we skip to the next valid PFN directly >> rather than skipping to the last invalid PFN so that the pfn++ in the > > hmm... Maybe this macro name makes you confused > > pfn = skip_to_last_invalid_pfn(pfn); > > how about skip_to_next_valid_pfn? > >> for () results in the next value. Can we replace the pfn++ there with >> a function calls that defaults to 'return pfn + 1', but does the skip >> for architectures that implement it? > > I am not sure I understand your question here. > With this patch, on !arm arches, skip_to_last_invalid_pfn is equal to (pfn), > and will be increased > when for{} loop continue. We only *skip* to the start pfn of next valid > region when > CONFIG_HAVE_MEMBLOCK and CONFIG_HAVE_ARCH_PFN_VALID(arm/arm64 supports > both). > What I am saying is that the loop in memmap_init_zone for (pfn = start_pfn; pfn < end_pfn; pfn++) { ... } should be replaced by something like for (pfn = start_pfn; pfn < end_pfn; pfn = next_valid_pfn(pfn)) where next_valid_pfn() is simply defined as static ulong next_valid_pfn(ulong pfn) { return pfn + 1; } by default, unless we do something special like you are proposing for ARM and arm64, in which case you provide a different implementation. That way, we no longer have to reason around the pfn++, and return an invalid pfn so that the ++ will produce a valid pfn
On 4/2/2018 3:53 PM, Ard Biesheuvel Wrote: > On 2 April 2018 at 09:49, Jia He <hejianet@gmail.com> wrote: >> >> On 4/2/2018 2:55 PM, Ard Biesheuvel Wrote: >>> On 2 April 2018 at 04:30, Jia He <hejianet@gmail.com> wrote: >>>> Commit b92df1de5d28 ("mm: page_alloc: skip over regions of invalid pfns >>>> where possible") optimized the loop in memmap_init_zone(). But it causes >>>> possible panic bug. So Daniel Vacek reverted it later. >>>> >>>> But as suggested by Daniel Vacek, it is fine to using memblock to skip >>>> gaps and finding next valid frame with CONFIG_HAVE_ARCH_PFN_VALID. >>>> >>>> On arm and arm64, memblock is used by default. But generic version of >>>> pfn_valid() is based on mem sections and memblock_next_valid_pfn() does >>>> not always return the next valid one but skips more resulting in some >>>> valid frames to be skipped (as if they were invalid). And that's why >>>> kernel was eventually crashing on some !arm machines. >>>> >>>> And as verified by Eugeniu Rosca, arm can benifit from commit >>>> b92df1de5d28. So remain the memblock_next_valid_pfn on arm{,64} and move >>>> the related codes to arm64 arch directory. >>>> >>>> Suggested-by: Daniel Vacek <neelx@redhat.com> >>>> Signed-off-by: Jia He <jia.he@hxt-semitech.com> >>> Hello Jia, >>> >>> Apologies for chiming in late. >> no problem, thanks for your comments ;-) >>> >>> If we are going to rearchitect this, I'd rather we change the loop in >>> memmap_init_zone() so that we skip to the next valid PFN directly >>> rather than skipping to the last invalid PFN so that the pfn++ in the >> hmm... Maybe this macro name makes you confused >> >> pfn = skip_to_last_invalid_pfn(pfn); >> >> how about skip_to_next_valid_pfn? >> >>> for () results in the next value. Can we replace the pfn++ there with >>> a function calls that defaults to 'return pfn + 1', but does the skip >>> for architectures that implement it? >> I am not sure I understand your question here. >> With this patch, on !arm arches, skip_to_last_invalid_pfn is equal to (pfn), >> and will be increased >> when for{} loop continue. We only *skip* to the start pfn of next valid >> region when >> CONFIG_HAVE_MEMBLOCK and CONFIG_HAVE_ARCH_PFN_VALID(arm/arm64 supports >> both). >> > What I am saying is that the loop in memmap_init_zone > > for (pfn = start_pfn; pfn < end_pfn; pfn++) { ... } > > should be replaced by something like > > for (pfn = start_pfn; pfn < end_pfn; pfn = next_valid_pfn(pfn)) > > where next_valid_pfn() is simply defined as > > static ulong next_valid_pfn(ulong pfn) > { > return pfn + 1; > } Hi Ard, Do you think a macro instead of simply fuction is better here? -- Cheer, Jia
On 4/2/2018 3:53 PM, Ard Biesheuvel Wrote: > On 2 April 2018 at 09:49, Jia He <hejianet@gmail.com> wrote: >> >> On 4/2/2018 2:55 PM, Ard Biesheuvel Wrote: >>> On 2 April 2018 at 04:30, Jia He <hejianet@gmail.com> wrote: >>>> Commit b92df1de5d28 ("mm: page_alloc: skip over regions of invalid pfns >>>> where possible") optimized the loop in memmap_init_zone(). But it causes >>>> possible panic bug. So Daniel Vacek reverted it later. >>>> >>>> But as suggested by Daniel Vacek, it is fine to using memblock to skip >>>> gaps and finding next valid frame with CONFIG_HAVE_ARCH_PFN_VALID. >>>> >>>> On arm and arm64, memblock is used by default. But generic version of >>>> pfn_valid() is based on mem sections and memblock_next_valid_pfn() does >>>> not always return the next valid one but skips more resulting in some >>>> valid frames to be skipped (as if they were invalid). And that's why >>>> kernel was eventually crashing on some !arm machines. >>>> >>>> And as verified by Eugeniu Rosca, arm can benifit from commit >>>> b92df1de5d28. So remain the memblock_next_valid_pfn on arm{,64} and move >>>> the related codes to arm64 arch directory. >>>> >>>> Suggested-by: Daniel Vacek <neelx@redhat.com> >>>> Signed-off-by: Jia He <jia.he@hxt-semitech.com> >>> Hello Jia, >>> >>> Apologies for chiming in late. >> no problem, thanks for your comments ;-) >>> >>> If we are going to rearchitect this, I'd rather we change the loop in >>> memmap_init_zone() so that we skip to the next valid PFN directly >>> rather than skipping to the last invalid PFN so that the pfn++ in the >> hmm... Maybe this macro name makes you confused >> >> pfn = skip_to_last_invalid_pfn(pfn); >> >> how about skip_to_next_valid_pfn? >> >>> for () results in the next value. Can we replace the pfn++ there with >>> a function calls that defaults to 'return pfn + 1', but does the skip >>> for architectures that implement it? >> I am not sure I understand your question here. >> With this patch, on !arm arches, skip_to_last_invalid_pfn is equal to (pfn), >> and will be increased >> when for{} loop continue. We only *skip* to the start pfn of next valid >> region when >> CONFIG_HAVE_MEMBLOCK and CONFIG_HAVE_ARCH_PFN_VALID(arm/arm64 supports >> both). >> > What I am saying is that the loop in memmap_init_zone > > for (pfn = start_pfn; pfn < end_pfn; pfn++) { ... } > > should be replaced by something like > > for (pfn = start_pfn; pfn < end_pfn; pfn = next_valid_pfn(pfn)) After further thinking, IMO, pfn = next_valid_pfn(pfn) might have impact on memmap_init_zone loop. e.g.context != MEMMAP_EARLY, pfn will not be checked by early_pfn_valid, thus It will change the memhotplug logic. So I would choose the old implementation: if (!early_pfn_valid(pfn)) { pfn = next_valid_pfn(pfn) - 1; continue; } Any comments? Thanks
diff --git a/arch/arm/include/asm/page.h b/arch/arm/include/asm/page.h index 4355f0e..489875c 100644 --- a/arch/arm/include/asm/page.h +++ b/arch/arm/include/asm/page.h @@ -158,6 +158,8 @@ typedef struct page *pgtable_t; #ifdef CONFIG_HAVE_ARCH_PFN_VALID extern int pfn_valid(unsigned long); +extern unsigned long memblock_next_valid_pfn(unsigned long pfn); +#define skip_to_last_invalid_pfn(pfn) (memblock_next_valid_pfn(pfn) - 1) #endif #include <asm/memory.h> diff --git a/arch/arm/mm/init.c b/arch/arm/mm/init.c index a1f11a7..0fb85ca 100644 --- a/arch/arm/mm/init.c +++ b/arch/arm/mm/init.c @@ -198,7 +198,36 @@ int pfn_valid(unsigned long pfn) return memblock_is_map_memory(__pfn_to_phys(pfn)); } EXPORT_SYMBOL(pfn_valid); -#endif + +/* HAVE_MEMBLOCK is always enabled on arm */ +unsigned long __init_memblock memblock_next_valid_pfn(unsigned long pfn) +{ + struct memblock_type *type = &memblock.memory; + unsigned int right = type->cnt; + unsigned int mid, left = 0; + phys_addr_t addr = PFN_PHYS(++pfn); + + do { + mid = (right + left) / 2; + + if (addr < type->regions[mid].base) + right = mid; + else if (addr >= (type->regions[mid].base + + type->regions[mid].size)) + left = mid + 1; + else { + /* addr is within the region, so pfn is valid */ + return pfn; + } + } while (left < right); + + if (right == type->cnt) + return -1UL; + else + return PHYS_PFN(type->regions[right].base); +} +EXPORT_SYMBOL(memblock_next_valid_pfn); +#endif /*CONFIG_HAVE_ARCH_PFN_VALID*/ #ifndef CONFIG_SPARSEMEM static void __init arm_memory_present(void) diff --git a/arch/arm64/include/asm/page.h b/arch/arm64/include/asm/page.h index 60d02c8..e57d3f2 100644 --- a/arch/arm64/include/asm/page.h +++ b/arch/arm64/include/asm/page.h @@ -39,6 +39,8 @@ typedef struct page *pgtable_t; #ifdef CONFIG_HAVE_ARCH_PFN_VALID extern int pfn_valid(unsigned long); +extern unsigned long memblock_next_valid_pfn(unsigned long pfn); +#define skip_to_last_invalid_pfn(pfn) (memblock_next_valid_pfn(pfn) - 1) #endif #include <asm/memory.h> diff --git a/arch/arm64/mm/init.c b/arch/arm64/mm/init.c index 00e7b90..13e43ff 100644 --- a/arch/arm64/mm/init.c +++ b/arch/arm64/mm/init.c @@ -290,7 +290,36 @@ int pfn_valid(unsigned long pfn) return memblock_is_map_memory(pfn << PAGE_SHIFT); } EXPORT_SYMBOL(pfn_valid); -#endif + +/* HAVE_MEMBLOCK is always enabled on arm64 */ +unsigned long __init_memblock memblock_next_valid_pfn(unsigned long pfn) +{ + struct memblock_type *type = &memblock.memory; + unsigned int right = type->cnt; + unsigned int mid, left = 0; + phys_addr_t addr = PFN_PHYS(++pfn); + + do { + mid = (right + left) / 2; + + if (addr < type->regions[mid].base) + right = mid; + else if (addr >= (type->regions[mid].base + + type->regions[mid].size)) + left = mid + 1; + else { + /* addr is within the region, so pfn is valid */ + return pfn; + } + } while (left < right); + + if (right == type->cnt) + return -1UL; + else + return PHYS_PFN(type->regions[right].base); +} +EXPORT_SYMBOL(memblock_next_valid_pfn); +#endif /*CONFIG_HAVE_ARCH_PFN_VALID*/ #ifndef CONFIG_SPARSEMEM static void __init arm64_memory_present(void) diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h index d797716..f9c0c46 100644 --- a/include/linux/mmzone.h +++ b/include/linux/mmzone.h @@ -1245,6 +1245,7 @@ static inline int pfn_valid(unsigned long pfn) return 0; return valid_section(__nr_to_section(pfn_to_section_nr(pfn))); } +#define skip_to_last_invalid_pfn(pfn) (pfn) #endif static inline int pfn_present(unsigned long pfn) diff --git a/mm/page_alloc.c b/mm/page_alloc.c index c19f5ac..30f7d76 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -5483,8 +5483,10 @@ void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone, if (context != MEMMAP_EARLY) goto not_early; - if (!early_pfn_valid(pfn)) + if (!early_pfn_valid(pfn)) { + pfn = skip_to_last_invalid_pfn(pfn); continue; + } if (!early_pfn_in_nid(pfn, nid)) continue; if (!update_defer_init(pgdat, pfn, end_pfn, &nr_initialised))
Commit b92df1de5d28 ("mm: page_alloc: skip over regions of invalid pfns where possible") optimized the loop in memmap_init_zone(). But it causes possible panic bug. So Daniel Vacek reverted it later. But as suggested by Daniel Vacek, it is fine to using memblock to skip gaps and finding next valid frame with CONFIG_HAVE_ARCH_PFN_VALID. On arm and arm64, memblock is used by default. But generic version of pfn_valid() is based on mem sections and memblock_next_valid_pfn() does not always return the next valid one but skips more resulting in some valid frames to be skipped (as if they were invalid). And that's why kernel was eventually crashing on some !arm machines. And as verified by Eugeniu Rosca, arm can benifit from commit b92df1de5d28. So remain the memblock_next_valid_pfn on arm{,64} and move the related codes to arm64 arch directory. Suggested-by: Daniel Vacek <neelx@redhat.com> Signed-off-by: Jia He <jia.he@hxt-semitech.com> --- arch/arm/include/asm/page.h | 2 ++ arch/arm/mm/init.c | 31 ++++++++++++++++++++++++++++++- arch/arm64/include/asm/page.h | 2 ++ arch/arm64/mm/init.c | 31 ++++++++++++++++++++++++++++++- include/linux/mmzone.h | 1 + mm/page_alloc.c | 4 +++- 6 files changed, 68 insertions(+), 3 deletions(-)