[patch 06/15] x86/tlb: Move __flush_tlb_global() out of line

From: Thomas Gleixner
Date: Sun Apr 19 2020 - 16:36:41 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 2nd step
move __flush_tlb_global() 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>
Cc: Juergen Gross <jgross@xxxxxxxx>
---
arch/x86/include/asm/paravirt.h | 1
arch/x86/include/asm/tlbflush.h | 38 +------------------------------------
arch/x86/kernel/paravirt.c | 9 --------
arch/x86/mm/tlb.c | 41 ++++++++++++++++++++++++++++++++++++++++
4 files changed, 44 insertions(+), 45 deletions(-)
--- a/arch/x86/include/asm/paravirt.h
+++ b/arch/x86/include/asm/paravirt.h
@@ -48,6 +48,7 @@ static inline void slow_down_io(void)
}

void native_flush_tlb_local(void);
+void native_flush_tlb_global(void);

static inline void __flush_tlb_local(void)
{
--- a/arch/x86/include/asm/tlbflush.h
+++ b/arch/x86/include/asm/tlbflush.h
@@ -141,11 +141,11 @@ static inline unsigned long build_cr3_no
}

void flush_tlb_local(void);
+void flush_tlb_global(void);

#ifdef CONFIG_PARAVIRT
#include <asm/paravirt.h>
#else
-#define __flush_tlb_global() __native_flush_tlb_global()
#define __flush_tlb_one_user(addr) __native_flush_tlb_one_user(addr)
#endif

@@ -372,40 +372,6 @@ static inline void invalidate_user_asid(
}

/*
- * flush everything
- */
-static inline void __native_flush_tlb_global(void)
-{
- unsigned long cr4, flags;
-
- if (static_cpu_has(X86_FEATURE_INVPCID)) {
- /*
- * Using INVPCID is considerably faster than a pair of writes
- * to CR4 sandwiched inside an IRQ flag save/restore.
- *
- * Note, this works with CR4.PCIDE=0 or 1.
- */
- invpcid_flush_all();
- return;
- }
-
- /*
- * Read-modify-write to CR4 - protect it from preemption and
- * from interrupts. (Use the raw variant because this code can
- * be called from deep inside debugging code.)
- */
- raw_local_irq_save(flags);
-
- cr4 = this_cpu_read(cpu_tlbstate.cr4);
- /* toggle PGE */
- native_write_cr4(cr4 ^ X86_CR4_PGE);
- /* write old PGE again and flush TLBs */
- native_write_cr4(cr4);
-
- raw_local_irq_restore(flags);
-}
-
-/*
* flush one page in the user mapping
*/
static inline void __native_flush_tlb_one_user(unsigned long addr)
@@ -439,7 +405,7 @@ static inline void __flush_tlb_all(void)
VM_WARN_ON_ONCE(preemptible());

if (boot_cpu_has(X86_FEATURE_PGE)) {
- __flush_tlb_global();
+ flush_tlb_global();
} else {
/*
* !PGE -> !PCID (setup_pcid()), thus every flush is total.
--- a/arch/x86/kernel/paravirt.c
+++ b/arch/x86/kernel/paravirt.c
@@ -160,15 +160,6 @@ unsigned paravirt_patch_insns(void *insn
return insn_len;
}

-/*
- * Global pages have to be flushed a bit differently. Not a real
- * performance problem because this does not happen often.
- */
-static void native_flush_tlb_global(void)
-{
- __native_flush_tlb_global();
-}
-
static void native_flush_tlb_one_user(unsigned long addr)
{
__native_flush_tlb_one_user(addr);
--- a/arch/x86/mm/tlb.c
+++ b/arch/x86/mm/tlb.c
@@ -23,6 +23,7 @@
#else
# define STATIC_NOPV static
# define __flush_tlb_local native_flush_tlb_local
+# define __flush_tlb_global native_flush_tlb_global
#endif

/*
@@ -891,6 +892,46 @@ unsigned long __get_current_cr3_fast(voi
EXPORT_SYMBOL_GPL(__get_current_cr3_fast);

/*
+ * Flush everything
+ */
+STATIC_NOPV void native_flush_tlb_global(void)
+{
+ unsigned long cr4, flags;
+
+ if (static_cpu_has(X86_FEATURE_INVPCID)) {
+ /*
+ * Using INVPCID is considerably faster than a pair of writes
+ * to CR4 sandwiched inside an IRQ flag save/restore.
+ *
+ * Note, this works with CR4.PCIDE=0 or 1.
+ */
+ invpcid_flush_all();
+ return;
+ }
+
+ /*
+ * Read-modify-write to CR4 - protect it from preemption and
+ * from interrupts. (Use the raw variant because this code can
+ * be called from deep inside debugging code.)
+ */
+ raw_local_irq_save(flags);
+
+ cr4 = this_cpu_read(cpu_tlbstate.cr4);
+ /* toggle PGE */
+ native_write_cr4(cr4 ^ X86_CR4_PGE);
+ /* write old PGE again and flush TLBs */
+ native_write_cr4(cr4);
+
+ raw_local_irq_restore(flags);
+}
+
+void flush_tlb_global(void)
+{
+ __flush_tlb_global();
+}
+EXPORT_SYMBOL_GPL(flush_tlb_global);
+
+/*
* Flush the entire current user mapping
*/
STATIC_NOPV void native_flush_tlb_local(void)