[patch 08/15] x86/tlb: Move __flush_tlb_one_kernel() out of line

From: Thomas Gleixner
Date: Sun Apr 19 2020 - 16:36:20 EST


cpu_tlbstate is exported because various TLB related functions need access
to it, but cpu_tlbstate is sensitive information which should only be
accessed by well contained kernel functions and not be directly exposed to
modules.

The various TLB flush functions need access to cpu_tlbstate. As forth step
move __flush_tlb_one_kernel() out of line and hide the native function. The
latter can be static when CONFIG_PARAVIRT is disabled.

Consolidate the name space while at it and remove the pointless extra
wrapper in the paravirt code.

No functional change.

Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
---
arch/x86/include/asm/pgtable_32.h | 2 -
arch/x86/include/asm/tlbflush.h | 41 --------------------------------------
arch/x86/mm/init_64.c | 2 -
arch/x86/mm/ioremap.c | 2 -
arch/x86/mm/kmmio.c | 2 -
arch/x86/mm/pat/set_memory.c | 2 -
arch/x86/mm/pgtable_32.c | 2 -
arch/x86/mm/tlb.c | 34 ++++++++++++++++++++++++++++++-
8 files changed, 40 insertions(+), 47 deletions(-)

--- a/arch/x86/include/asm/pgtable_32.h
+++ b/arch/x86/include/asm/pgtable_32.h
@@ -60,7 +60,7 @@ void sync_initial_page_table(void);
#define kpte_clear_flush(ptep, vaddr) \
do { \
pte_clear(&init_mm, (vaddr), (ptep)); \
- __flush_tlb_one_kernel((vaddr)); \
+ flush_tlb_one_kernel((vaddr)); \
} while (0)

#endif /* !__ASSEMBLY__ */
--- a/arch/x86/include/asm/tlbflush.h
+++ b/arch/x86/include/asm/tlbflush.h
@@ -143,6 +143,7 @@ static inline unsigned long build_cr3_no
void flush_tlb_local(void);
void flush_tlb_global(void);
void flush_tlb_one_user(unsigned long addr);
+void flush_tlb_one_kernel(unsigned long addr);

#ifdef CONFIG_PARAVIRT
#include <asm/paravirt.h>
@@ -318,14 +319,6 @@ static inline void cr4_clear_bits(unsign
}

/*
- * Mark all other ASIDs as invalid, preserves the current.
- */
-static inline void invalidate_other_asid(void)
-{
- this_cpu_write(cpu_tlbstate.invalidate_other, true);
-}
-
-/*
* Save some of cr4 feature set we're using (e.g. Pentium 4MB
* enable and PPro Global page enable), so that any CPU's that boot
* up after us can get the correct flags. This should only be used
@@ -365,38 +358,6 @@ static inline void __flush_tlb_all(void)
}
}

-/*
- * flush one page in the kernel mapping
- */
-static inline void __flush_tlb_one_kernel(unsigned long addr)
-{
- count_vm_tlb_event(NR_TLB_LOCAL_FLUSH_ONE);
-
- /*
- * If PTI is off, then __flush_tlb_one_user() is just INVLPG or its
- * paravirt equivalent. Even with PCID, this is sufficient: we only
- * use PCID if we also use global PTEs for the kernel mapping, and
- * INVLPG flushes global translations across all address spaces.
- *
- * If PTI is on, then the kernel is mapped with non-global PTEs, and
- * __flush_tlb_one_user() will flush the given address for the current
- * kernel address space and for its usermode counterpart, but it does
- * not flush it for other address spaces.
- */
- flush_tlb_one_user(addr);
-
- if (!static_cpu_has(X86_FEATURE_PTI))
- return;
-
- /*
- * See above. We need to propagate the flush to all other address
- * spaces. In principle, we only need to propagate it to kernelmode
- * address spaces, but the extra bookkeeping we would need is not
- * worth it.
- */
- invalidate_other_asid();
-}
-
#define TLB_FLUSH_ALL -1UL

/*
--- a/arch/x86/mm/init_64.c
+++ b/arch/x86/mm/init_64.c
@@ -298,7 +298,7 @@ static void __set_pte_vaddr(pud_t *pud,
* It's enough to flush this one mapping.
* (PGE mappings get flushed as well)
*/
- __flush_tlb_one_kernel(vaddr);
+ flush_tlb_one_kernel(vaddr);
}

void set_pte_vaddr_p4d(p4d_t *p4d_page, unsigned long vaddr, pte_t new_pte)
--- a/arch/x86/mm/ioremap.c
+++ b/arch/x86/mm/ioremap.c
@@ -889,5 +889,5 @@ void __init __early_set_fixmap(enum fixe
set_pte(pte, pfn_pte(phys >> PAGE_SHIFT, flags));
else
pte_clear(&init_mm, addr, pte);
- __flush_tlb_one_kernel(addr);
+ flush_tlb_one_kernel(addr);
}
--- a/arch/x86/mm/kmmio.c
+++ b/arch/x86/mm/kmmio.c
@@ -173,7 +173,7 @@ static int clear_page_presence(struct km
return -1;
}

- __flush_tlb_one_kernel(f->addr);
+ flush_tlb_one_kernel(f->addr);
return 0;
}

--- a/arch/x86/mm/pat/set_memory.c
+++ b/arch/x86/mm/pat/set_memory.c
@@ -340,7 +340,7 @@ static void __cpa_flush_tlb(void *data)
unsigned int i;

for (i = 0; i < cpa->numpages; i++)
- __flush_tlb_one_kernel(fix_addr(__cpa_addr(cpa, i)));
+ flush_tlb_one_kernel(fix_addr(__cpa_addr(cpa, i)));
}

static void cpa_flush(struct cpa_data *data, int cache)
--- a/arch/x86/mm/pgtable_32.c
+++ b/arch/x86/mm/pgtable_32.c
@@ -64,7 +64,7 @@ void set_pte_vaddr(unsigned long vaddr,
* It's enough to flush this one mapping.
* (PGE mappings get flushed as well)
*/
- __flush_tlb_one_kernel(vaddr);
+ flush_tlb_one_kernel(vaddr);
}

unsigned long __FIXADDR_TOP = 0xfffff000;
--- a/arch/x86/mm/tlb.c
+++ b/arch/x86/mm/tlb.c
@@ -876,7 +876,7 @@ static void do_kernel_range_flush(void *

/* flush range by one by one 'invlpg' */
for (addr = f->start; addr < f->end; addr += PAGE_SIZE)
- __flush_tlb_one_kernel(addr);
+ flush_tlb_one_kernel(addr);
}

void flush_tlb_kernel_range(unsigned long start, unsigned long end)
@@ -919,6 +919,38 @@ unsigned long __get_current_cr3_fast(voi
EXPORT_SYMBOL_GPL(__get_current_cr3_fast);

/*
+ * Flush one page in the kernel mapping
+ */
+void flush_tlb_one_kernel(unsigned long addr)
+{
+ count_vm_tlb_event(NR_TLB_LOCAL_FLUSH_ONE);
+
+ /*
+ * If PTI is off, then __flush_tlb_one_user() is just INVLPG or its
+ * paravirt equivalent. Even with PCID, this is sufficient: we only
+ * use PCID if we also use global PTEs for the kernel mapping, and
+ * INVLPG flushes global translations across all address spaces.
+ *
+ * If PTI is on, then the kernel is mapped with non-global PTEs, and
+ * __flush_tlb_one_user() will flush the given address for the current
+ * kernel address space and for its usermode counterpart, but it does
+ * not flush it for other address spaces.
+ */
+ flush_tlb_one_user(addr);
+
+ if (!static_cpu_has(X86_FEATURE_PTI))
+ return;
+
+ /*
+ * See above. We need to propagate the flush to all other address
+ * spaces. In principle, we only need to propagate it to kernelmode
+ * address spaces, but the extra bookkeeping we would need is not
+ * worth it.
+ */
+ this_cpu_write(cpu_tlbstate.invalidate_other, true);
+}
+
+/*
* Flush one page in the user mapping
*/
STATIC_NOPV void native_flush_tlb_one_user(unsigned long addr)