From 1f9cc3cb6a27521edfe0a21abf97d2bb11c4d237 Mon Sep 17 00:00:00 2001 From: Robin Holt Date: Fri, 23 Apr 2010 10:36:22 -0500 Subject: x86, pat: Update the page flags for memtype atomically instead of using memtype_lock While testing an application using the xpmem (out of kernel) driver, we noticed a significant page fault rate reduction of x86_64 with respect to ia64. For one test running with 32 cpus, one thread per cpu, it took 01:08 for each of the threads to vm_insert_pfn 2GB worth of pages. For the same test running on 256 cpus, one thread per cpu, it took 14:48 to vm_insert_pfn 2 GB worth of pages. The slowdown was tracked to lookup_memtype which acquires the spinlock memtype_lock. This heavily contended lock was slowing down vm_insert_pfn(). With the cmpxchg on page->flags method, both the 32 cpu and 256 cpu cases take approx 00:01.3 seconds to complete. Signed-off-by: Robin Holt LKML-Reference: <20100423153627.751194346@gulag1.americas.sgi.com> Cc: Venkatesh Pallipadi Cc: Rafael Wysocki Reviewed-by: Suresh Siddha Signed-off-by: H. Peter Anvin --- arch/x86/mm/pat.c | 8 -------- 1 file changed, 8 deletions(-) (limited to 'arch/x86/mm') diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c index 951011166ef5..501fc60e5e4d 100644 --- a/arch/x86/mm/pat.c +++ b/arch/x86/mm/pat.c @@ -190,8 +190,6 @@ static int pat_pagerange_is_ram(unsigned long start, unsigned long end) * Here we do two pass: * - Find the memtype of all the pages in the range, look for any conflicts * - In case of no conflicts, set the new memtype for pages in the range - * - * Caller must hold memtype_lock for atomicity. */ static int reserve_ram_pages_type(u64 start, u64 end, unsigned long req_type, unsigned long *new_type) @@ -297,9 +295,7 @@ int reserve_memtype(u64 start, u64 end, unsigned long req_type, is_range_ram = pat_pagerange_is_ram(start, end); if (is_range_ram == 1) { - spin_lock(&memtype_lock); err = reserve_ram_pages_type(start, end, req_type, new_type); - spin_unlock(&memtype_lock); return err; } else if (is_range_ram < 0) { @@ -351,9 +347,7 @@ int free_memtype(u64 start, u64 end) is_range_ram = pat_pagerange_is_ram(start, end); if (is_range_ram == 1) { - spin_lock(&memtype_lock); err = free_ram_pages_type(start, end); - spin_unlock(&memtype_lock); return err; } else if (is_range_ram < 0) { @@ -394,10 +388,8 @@ static unsigned long lookup_memtype(u64 paddr) if (pat_pagerange_is_ram(paddr, paddr + PAGE_SIZE)) { struct page *page; - spin_lock(&memtype_lock); page = pfn_to_page(paddr >> PAGE_SHIFT); rettype = get_page_memtype(page); - spin_unlock(&memtype_lock); /* * -1 from get_page_memtype() implies RAM page is in its * default state and not reserved, and hence of type WB -- cgit v1.2.2