xiaomi-sm8450/android_kernel_xiaomi_sm8450
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

arch, mm: Remove tlb_fast_mode()

Browse Source 
Since the introduction of preemptible mmu_gather TLB fast mode has been
broken. TLB fast mode relies on there being absolutely no concurrency;
it frees pages first and invalidates TLBs later.

However now we can get concurrency and stuff goes *bang*.

This patch removes all tlb_fast_mode() code; it was found the better
option vs trying to patch the hole by entangling tlb invalidation with
the scheduler.

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Tony Luck <tony.luck@intel.com>
Reported-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This commit is contained in:
Peter Zijlstra
2013-06-05 12:26:50 +02:00
committed by Linus Torvalds
parent 2c95523c0f
commit 29eb77825c
4 changed files with 13 additions and 81 deletions
Download Patch File Download Diff File Expand all files Collapse all files

41
arch/ia64/include/asm/tlb.h
View File

@@ -46,12 +46,6 @@
#include <asm/tlbflush.h>
#include <asm/machvec.h>
#ifdef CONFIG_SMP
# define tlb_fast_mode(tlb) ((tlb)->nr == ~0U)
#else
# define tlb_fast_mode(tlb) (1)
#endif
/*
* If we can't allocate a page to make a big batch of page pointers
* to work on, then just handle a few from the on-stack structure.
@@ -60,7 +54,7 @@
struct mmu_gather {
struct mm_struct *mm;
unsigned int nr; /* == ~0U => fast mode */
unsigned int nr;
unsigned int max;
unsigned char fullmm; /* non-zero means full mm flush */
unsigned char need_flush; /* really unmapped some PTEs? */
@@ -103,6 +97,7 @@ extern struct ia64_tr_entry *ia64_idtrs[NR_CPUS];
static inline void
ia64_tlb_flush_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long end)
{
unsigned long i;
unsigned int nr;
if (!tlb->need_flush)
@@ -141,13 +136,11 @@ ia64_tlb_flush_mmu (struct mmu_gather *tlb, unsigned long start, unsigned long e
/* lastly, release the freed pages */
nr = tlb->nr;
if (!tlb_fast_mode(tlb)) {
unsigned long i;
tlb->nr = 0;
tlb->start_addr = ~0UL;
for (i = 0; i < nr; ++i)
free_page_and_swap_cache(tlb->pages[i]);
}
tlb->nr = 0;
tlb->start_addr = ~0UL;
for (i = 0; i < nr; ++i)
free_page_and_swap_cache(tlb->pages[i]);
}
static inline void __tlb_alloc_page(struct mmu_gather *tlb)
@@ -167,20 +160,7 @@ tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, unsigned int full_m
tlb->mm = mm;
tlb->max = ARRAY_SIZE(tlb->local);
tlb->pages = tlb->local;
/*
* Use fast mode if only 1 CPU is online.
*
* It would be tempting to turn on fast-mode for full_mm_flush as well. But this
* doesn't work because of speculative accesses and software prefetching: the page
* table of "mm" may (and usually is) the currently active page table and even
* though the kernel won't do any user-space accesses during the TLB shoot down, a
* compiler might use speculation or lfetch.fault on what happens to be a valid
* user-space address. This in turn could trigger a TLB miss fault (or a VHPT
* walk) and re-insert a TLB entry we just removed. Slow mode avoids such
* problems. (We could make fast-mode work by switching the current task to a
* different "mm" during the shootdown.) --davidm 08/02/2002
*/
tlb->nr = (num_online_cpus() == 1) ? ~0U : 0;
tlb->nr = 0;
tlb->fullmm = full_mm_flush;
tlb->start_addr = ~0UL;
}
@@ -214,11 +194,6 @@ static inline int __tlb_remove_page(struct mmu_gather *tlb, struct page *page)
{
tlb->need_flush = 1;
if (tlb_fast_mode(tlb)) {
free_page_and_swap_cache(page);
return 1; /* avoid calling tlb_flush_mmu */
}
if (!tlb->nr && tlb->pages == tlb->local)
__tlb_alloc_page(tlb);