[v3,1/7] mm: Add functions to track page directory modifications
diff mbox series

Message ID 20200515140023.25469-2-joro@8bytes.org
State New
Headers show
Series
  • mm: Get rid of vmalloc_sync_(un)mappings()
Related show

Commit Message

Joerg Roedel May 15, 2020, 2 p.m. UTC
From: Joerg Roedel <jroedel@suse.de>

Add page-table allocation functions which will keep track of changed
directory entries. They are needed for new PGD, P4D, PUD, and PMD
entries and will be used in vmalloc and ioremap code to decide whether
any changes in the kernel mappings need to be synchronized between
page-tables in the system.

Signed-off-by: Joerg Roedel <jroedel@suse.de>
Acked-by: Andy Lutomirski <luto@kernel.org>
---
 include/asm-generic/5level-fixup.h |  5 ++--
 include/asm-generic/pgtable.h      | 23 +++++++++++++++
 include/linux/mm.h                 | 46 ++++++++++++++++++++++++++++++
 3 files changed, 72 insertions(+), 2 deletions(-)

Comments

Catalin Marinas June 5, 2020, 10:08 a.m. UTC | #1
Hi Joerg,

On Fri, May 15, 2020 at 04:00:17PM +0200, Joerg Roedel wrote:
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index 5a323422d783..022fe682af9e 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -2078,13 +2078,54 @@ static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
>  	return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
>  		NULL : pud_offset(p4d, address);
>  }
> +
> +static inline p4d_t *p4d_alloc_track(struct mm_struct *mm, pgd_t *pgd,
> +				     unsigned long address,
> +				     pgtbl_mod_mask *mod_mask)
> +
> +{
> +	if (unlikely(pgd_none(*pgd))) {
> +		if (__p4d_alloc(mm, pgd, address))
> +			return NULL;
> +		*mod_mask |= PGTBL_PGD_MODIFIED;
> +	}
> +
> +	return p4d_offset(pgd, address);
> +}
> +
>  #endif /* !__ARCH_HAS_5LEVEL_HACK */
>  
> +static inline pud_t *pud_alloc_track(struct mm_struct *mm, p4d_t *p4d,
> +				     unsigned long address,
> +				     pgtbl_mod_mask *mod_mask)
> +{
> +	if (unlikely(p4d_none(*p4d))) {
> +		if (__pud_alloc(mm, p4d, address))
> +			return NULL;
> +		*mod_mask |= PGTBL_P4D_MODIFIED;
> +	}
> +
> +	return pud_offset(p4d, address);
> +}

This patch causes a kernel panic on arm64 (and possibly powerpc, I
haven't tried). arm64 still uses the 5level-fixup.h and pud_alloc()
checks for the empty p4d with pgd_none() instead of p4d_none().

The patch below fixes it:

-----------------------8<-----------------------------------
From 8fb5f4c1e0983bb501c8ec71f9bb9dd344b80e87 Mon Sep 17 00:00:00 2001
From: Catalin Marinas <catalin.marinas@arm.com>
Date: Fri, 5 Jun 2020 10:56:03 +0100
Subject: [PATCH] mm: Use pgd_none() in pud_alloc_track() with
 __ARCH_HAS_5LEVEL_HACK

Commit d8626138009b ("mm: add functions to track page directory
modifications") introduced the pud_alloc_track() function checking for
an empty p4d using p4d_none(). However, when __ARCH_HAS_5LEVEL_HACK is
defined, the pud_alloc() counterpart checks for an empty p4d using
pgd_none(). Since p4d_none() is always 0 in this case, no pud would be
allocated and the kernel panics during boot on arm64 (at least).

Until all architectures are moved away from the 5level-fixup.h, define a
pud_alloc_track() that matches the __ARCH_HAS_5LEVEL_HACK pud_alloc().

Fixes: d8626138009b ("mm: add functions to track page directory modifications")
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: Joerg Roedel <jroedel@suse.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
---
 include/linux/mm.h | 19 +++++++++++++++++--
 1 file changed, 17 insertions(+), 2 deletions(-)

diff --git a/include/linux/mm.h b/include/linux/mm.h
index fda41eb7f1c8..9d3761a1fad5 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -2106,8 +2106,6 @@ static inline p4d_t *p4d_alloc_track(struct mm_struct *mm, pgd_t *pgd,
 	return p4d_offset(pgd, address);
 }
 
-#endif /* !__ARCH_HAS_5LEVEL_HACK */
-
 static inline pud_t *pud_alloc_track(struct mm_struct *mm, p4d_t *p4d,
 				     unsigned long address,
 				     pgtbl_mod_mask *mod_mask)
@@ -2121,6 +2119,23 @@ static inline pud_t *pud_alloc_track(struct mm_struct *mm, p4d_t *p4d,
 	return pud_offset(p4d, address);
 }
 
+#else /* __ARCH_HAS_5LEVEL_HACK */
+
+static inline pud_t *pud_alloc_track(struct mm_struct *mm, p4d_t *p4d,
+				     unsigned long address,
+				     pgtbl_mod_mask *mod_mask)
+{
+	if (unlikely(pgd_none(*p4d))) {
+		if (__pud_alloc(mm, p4d, address))
+			return NULL;
+		*mod_mask |= PGTBL_P4D_MODIFIED;
+	}
+
+	return pud_offset(p4d, address);
+}
+
+#endif /* !__ARCH_HAS_5LEVEL_HACK */
+
 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
 {
 	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
Catalin Marinas June 5, 2020, 11:46 a.m. UTC | #2
On Fri, Jun 05, 2020 at 11:08:13AM +0100, Catalin Marinas wrote:
> This patch causes a kernel panic on arm64 (and possibly powerpc, I
> haven't tried). arm64 still uses the 5level-fixup.h and pud_alloc()
> checks for the empty p4d with pgd_none() instead of p4d_none().

Ah, should have checked the list first
(https://lore.kernel.org/linux-mm/20200604074446.23944-1-joro@8bytes.org/).

Please ignore my patch then.
Andrew Morton June 6, 2020, 1:01 a.m. UTC | #3
On Fri, 5 Jun 2020 12:46:55 +0100 Catalin Marinas <catalin.marinas@arm.com> wrote:

> On Fri, Jun 05, 2020 at 11:08:13AM +0100, Catalin Marinas wrote:
> > This patch causes a kernel panic on arm64 (and possibly powerpc, I
> > haven't tried). arm64 still uses the 5level-fixup.h and pud_alloc()
> > checks for the empty p4d with pgd_none() instead of p4d_none().
> 
> Ah, should have checked the list first
> (https://lore.kernel.org/linux-mm/20200604074446.23944-1-joro@8bytes.org/).
> 

Sorry about that.

Can you confirm that the merge of Mike's 5level_hack zappage has fixed
the arm64 wontboot?
Catalin Marinas June 6, 2020, 12:50 p.m. UTC | #4
On Fri, Jun 05, 2020 at 06:01:16PM -0700, Andrew Morton wrote:
> On Fri, 5 Jun 2020 12:46:55 +0100 Catalin Marinas <catalin.marinas@arm.com> wrote:
> > On Fri, Jun 05, 2020 at 11:08:13AM +0100, Catalin Marinas wrote:
> > > This patch causes a kernel panic on arm64 (and possibly powerpc, I
> > > haven't tried). arm64 still uses the 5level-fixup.h and pud_alloc()
> > > checks for the empty p4d with pgd_none() instead of p4d_none().
> > 
> > Ah, should have checked the list first
> > (https://lore.kernel.org/linux-mm/20200604074446.23944-1-joro@8bytes.org/).
> 
> Can you confirm that the merge of Mike's 5level_hack zappage has fixed
> the arm64 wontboot?

Yes, it has. Mainline is booting again on arm64.

Thanks.

Patch
diff mbox series

diff --git a/include/asm-generic/5level-fixup.h b/include/asm-generic/5level-fixup.h
index 4c74b1c1d13b..58046ddc08d0 100644
--- a/include/asm-generic/5level-fixup.h
+++ b/include/asm-generic/5level-fixup.h
@@ -17,8 +17,9 @@ 
 	((unlikely(pgd_none(*(p4d))) && __pud_alloc(mm, p4d, address)) ? \
 		NULL : pud_offset(p4d, address))
 
-#define p4d_alloc(mm, pgd, address)	(pgd)
-#define p4d_offset(pgd, start)		(pgd)
+#define p4d_alloc(mm, pgd, address)		(pgd)
+#define p4d_alloc_track(mm, pgd, address, mask)	(pgd)
+#define p4d_offset(pgd, start)			(pgd)
 
 #ifndef __ASSEMBLY__
 static inline int p4d_none(p4d_t p4d)
diff --git a/include/asm-generic/pgtable.h b/include/asm-generic/pgtable.h
index 329b8c8ca703..bf1418ae91a2 100644
--- a/include/asm-generic/pgtable.h
+++ b/include/asm-generic/pgtable.h
@@ -1209,6 +1209,29 @@  static inline bool arch_has_pfn_modify_check(void)
 # define PAGE_KERNEL_EXEC PAGE_KERNEL
 #endif
 
+/*
+ * Page Table Modification bits for pgtbl_mod_mask.
+ *
+ * These are used by the p?d_alloc_track*() set of functions an in the generic
+ * vmalloc/ioremap code to track at which page-table levels entries have been
+ * modified. Based on that the code can better decide when vmalloc and ioremap
+ * mapping changes need to be synchronized to other page-tables in the system.
+ */
+#define		__PGTBL_PGD_MODIFIED	0
+#define		__PGTBL_P4D_MODIFIED	1
+#define		__PGTBL_PUD_MODIFIED	2
+#define		__PGTBL_PMD_MODIFIED	3
+#define		__PGTBL_PTE_MODIFIED	4
+
+#define		PGTBL_PGD_MODIFIED	BIT(__PGTBL_PGD_MODIFIED)
+#define		PGTBL_P4D_MODIFIED	BIT(__PGTBL_P4D_MODIFIED)
+#define		PGTBL_PUD_MODIFIED	BIT(__PGTBL_PUD_MODIFIED)
+#define		PGTBL_PMD_MODIFIED	BIT(__PGTBL_PMD_MODIFIED)
+#define		PGTBL_PTE_MODIFIED	BIT(__PGTBL_PTE_MODIFIED)
+
+/* Page-Table Modification Mask */
+typedef unsigned int pgtbl_mod_mask;
+
 #endif /* !__ASSEMBLY__ */
 
 #ifndef io_remap_pfn_range
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 5a323422d783..022fe682af9e 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -2078,13 +2078,54 @@  static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
 	return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
 		NULL : pud_offset(p4d, address);
 }
+
+static inline p4d_t *p4d_alloc_track(struct mm_struct *mm, pgd_t *pgd,
+				     unsigned long address,
+				     pgtbl_mod_mask *mod_mask)
+
+{
+	if (unlikely(pgd_none(*pgd))) {
+		if (__p4d_alloc(mm, pgd, address))
+			return NULL;
+		*mod_mask |= PGTBL_PGD_MODIFIED;
+	}
+
+	return p4d_offset(pgd, address);
+}
+
 #endif /* !__ARCH_HAS_5LEVEL_HACK */
 
+static inline pud_t *pud_alloc_track(struct mm_struct *mm, p4d_t *p4d,
+				     unsigned long address,
+				     pgtbl_mod_mask *mod_mask)
+{
+	if (unlikely(p4d_none(*p4d))) {
+		if (__pud_alloc(mm, p4d, address))
+			return NULL;
+		*mod_mask |= PGTBL_P4D_MODIFIED;
+	}
+
+	return pud_offset(p4d, address);
+}
+
 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
 {
 	return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
 		NULL: pmd_offset(pud, address);
 }
+
+static inline pmd_t *pmd_alloc_track(struct mm_struct *mm, pud_t *pud,
+				     unsigned long address,
+				     pgtbl_mod_mask *mod_mask)
+{
+	if (unlikely(pud_none(*pud))) {
+		if (__pmd_alloc(mm, pud, address))
+			return NULL;
+		*mod_mask |= PGTBL_PUD_MODIFIED;
+	}
+
+	return pmd_offset(pud, address);
+}
 #endif /* CONFIG_MMU */
 
 #if USE_SPLIT_PTE_PTLOCKS
@@ -2200,6 +2241,11 @@  static inline void pgtable_pte_page_dtor(struct page *page)
 	((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd))? \
 		NULL: pte_offset_kernel(pmd, address))
 
+#define pte_alloc_kernel_track(pmd, address, mask)			\
+	((unlikely(pmd_none(*(pmd))) &&					\
+	  (__pte_alloc_kernel(pmd) || ({*(mask)|=PGTBL_PMD_MODIFIED;0;})))?\
+		NULL: pte_offset_kernel(pmd, address))
+
 #if USE_SPLIT_PMD_PTLOCKS
 
 static struct page *pmd_to_page(pmd_t *pmd)