Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit 719ea79e authored by Chris Metcalf's avatar Chris Metcalf
Browse files

arch/tile: fix up locking in pgtable.c slightly 



We should be holding the init_mm.page_table_lock in shatter_huge_page()
since we are modifying the kernel page tables.  Then, only if we are
walking the other root page tables to update them, do we want to take
the pgd_lock.

Add a comment about taking the pgd_lock that we always do it with
interrupts disabled and therefore are not at risk from the tlbflush
IPI deadlock as is seen on x86.

Signed-off-by: default avatarChris Metcalf <cmetcalf@tilera.com>
parent 5f220704

arch/tile/mm/pgtable.c

+12 −10
Original line number Diff line number Diff line
@@ -177,14 +177,10 @@ void shatter_huge_page(unsigned long addr) 
	if (!pmd_huge_page(*pmd))

		return;



	/*

	 * Grab the pgd_lock, since we may need it to walk the pgd_list,

	 * and since we need some kind of lock here to avoid races.

	 */

	spin_lock_irqsave(&pgd_lock, flags);

	spin_lock_irqsave(&init_mm.page_table_lock, flags);

	if (!pmd_huge_page(*pmd)) {

		/* Lost the race to convert the huge page. */

		spin_unlock_irqrestore(&pgd_lock, flags);

		spin_unlock_irqrestore(&init_mm.page_table_lock, flags);

		return;

	}
@@ -194,6 +190,7 @@ void shatter_huge_page(unsigned long addr) 


#ifdef __PAGETABLE_PMD_FOLDED

	/* Walk every pgd on the system and update the pmd there. */

	spin_lock(&pgd_lock);

	list_for_each(pos, &pgd_list) {

		pmd_t *copy_pmd;

		pgd = list_to_pgd(pos) + pgd_index(addr);
@@ -201,6 +198,7 @@ void shatter_huge_page(unsigned long addr) 
		copy_pmd = pmd_offset(pud, addr);

		__set_pmd(copy_pmd, *pmd);

	}

	spin_unlock(&pgd_lock);

#endif



	/* Tell every cpu to notice the change. */
@@ -208,7 +206,7 @@ void shatter_huge_page(unsigned long addr) 
		     cpu_possible_mask, NULL, 0);



	/* Hold the lock until the TLB flush is finished to avoid races. */

	spin_unlock_irqrestore(&pgd_lock, flags);

	spin_unlock_irqrestore(&init_mm.page_table_lock, flags);

}



/*
@@ -217,9 +215,13 @@ void shatter_huge_page(unsigned long addr) 
 * against pageattr.c; it is the unique case in which a valid change

 * of kernel pagetables can't be lazily synchronized by vmalloc faults.

 * vmalloc faults work because attached pagetables are never freed.

 * The locking scheme was chosen on the basis of manfred's

 * recommendations and having no core impact whatsoever.

 * -- wli

 *

 * The lock is always taken with interrupts disabled, unlike on x86

 * and other platforms, because we need to take the lock in

 * shatter_huge_page(), which may be called from an interrupt context.

 * We are not at risk from the tlbflush IPI deadlock that was seen on

 * x86, since we use the flush_remote() API to have the hypervisor do

 * the TLB flushes regardless of irq disabling.

 */

DEFINE_SPINLOCK(pgd_lock);

LIST_HEAD(pgd_list);