Messages in this thread | | | Date | Thu, 5 Nov 2015 10:50:33 +0200 | From | "Kirill A. Shutemov" <> | Subject | Re: [PATCH V2] mm: fix kernel crash in khugepaged thread |
| |
On Thu, Nov 05, 2015 at 09:12:34AM +0100, Vlastimil Babka wrote: > On 10/29/2015 01:35 AM, Kirill A. Shutemov wrote: > >> @@ -2605,9 +2603,9 @@ out_unmap: > >> /* collapse_huge_page will return with the mmap_sem released */ > >> collapse_huge_page(mm, address, hpage, vma, node); > >> } > >> -out: > >> - trace_mm_khugepaged_scan_pmd(mm, page_to_pfn(page), writable, referenced, > >> - none_or_zero, result, unmapped); > >> + trace_mm_khugepaged_scan_pmd(mm, pte_present(pteval) ? > >> + pte_pfn(pteval) : -1, writable, referenced, > >> + none_or_zero, result, unmapped); > > > > maybe passing down pte instead of pfn? > > Maybe just pass the page, and have tracepoint's fast assign check for !NULL and > do page_to_pfn itself? That way the complexity and overhead is only in the > tracepoint and when enabled.
Agreed.
-- Kirill A. Shutemov
| |