proc: mm: export PTE sizes directly in smaps
Subsystems (5)
PATCH v2 1/2 proc: mm: export PTE sizes directly in smaps
/proc/$pid/smaps has a number of fields that are intended to imply the kinds of PTEs used to map memory. "AnonHugePages" obviously tells you how many PMDs are being used. "MMUPageSize" along with the "Hugetlb" fields tells you how many PTEs you have for a huge page.
The current mechanisms work fine when we have one or two page sizes. But, they start to get a bit muddled when we mix page sizes inside one VMA. For instance, the DAX folks were proposing adding a set of fields like:
DevicePages:
DeviceHugePages:
DeviceGiganticPages:
DeviceGinormousPages:
to unmuddle things when page sizes get mixed. That's fine, but it does require userspace know the mapping from our various arbitrary names to hardware page sizes on each architecture and kernel configuration. That seems rather suboptimal.
What folks really want is to know how much memory is mapped with each page size. How about we just do *that* instead?
Patch attached. Seems harmless enough. Seems to compile on a bunch of random architectures. Makes smaps look like this:
Private_Hugetlb: 0 kB Swap: 0 kB SwapPss: 0 kB KernelPageSize: 4 kB MMUPageSize: 4 kB Locked: 0 kB
Ptes@4kB: 32 kB
Ptes@2MB: 2048 kB
The format I used here should be unlikely to break smaps parsers unless they're looking for "kB" and now match the 'Ptes@4kB' instead of the one at the end of the line.
Note: hugetlbfs PTEs are unusual. We can have more than one "pte_t" for each hugetlbfs "page". arm64 has this configuration, and probably others. The code should now handle when an hstate's size is not equal to one of the page table entry sizes. For instance, it assumes that hstates between PMD_SIZE and PUD_SIZE are made up of multiple PMDs and prints them as such.
I've tested this on x86 with normal 4k ptes, anonymous huge pages, 1G hugetlbfs and 2M hugetlbfs pages.
1. I'd like to thank Dan Williams for showing me a mirror as I complained about the bozo that introduced 'AnonHugePages'.
[Fan]
Rebase the original patch from Dave Hansen by fixing a couple of compile issues.
--- Documentation/filesystems/proc.txt | 6 +++ fs/proc/task_mmu.c | 106 ++++++++++++++++++++++++++++++++++++- mm/hugetlb.c | 11 ++++ 3 files changed, 121 insertions(+), 2 deletions(-)
Documentation/filesystems/proc.txt +6 -0
@@ -423,6 +423,9 @@ SwapPss: 0 kB KernelPageSize: 4 kB MMUPageSize: 4 kB Locked: 0 kB +Ptes@4kB: 4 kB +Ptes@2MB: 8192 kB + VmFlags: rd ex mr mw me dw the first of these lines shows the same information as is displayed for the
@@ -460,6 +463,9 @@ replaced by copy-on-write) part of the underlying shmem object out on swap. "SwapPss" shows proportional swap share of this mapping. Unlike "Swap", this does not take into account swapped out page of underlying shmem objects. "Locked" indicates whether the mapping is locked in memory or not. +"Ptes@..." lines show how many page table entries are currently in place and +pointing to memory. There is an entry for each size present in the hardware +page tables for this mapping. "VmFlags" field deserves a separate description. This member represents the kernel flags associated with the particular virtual memory area in two letter encoded
fs/proc/task_mmu.c +104 -2
151 lines not on this page · raw message
mm/hugetlb.c +11 -0
@@ -2836,6 +2836,17 @@ void __init hugetlb_add_hstate(unsigned int order) huge_page_size(h)/1024); parsed_hstate = h; + +/* +* PGD_SIZE isn't widely made available by architecures, +* so use PUD_SIZE*PTRS_PER_PUD as a substitute. +* +* Check for sizes that might be mapped by a PGD. There +* are none of these known today, but be on the lookout. +* If this trips, we will need to update the mss->rss_* +* code in fs/proc/task_mmu.c. +*/ +WARN_ON_ONCE ((PAGE_SIZE << order) >=PUD_SIZE *PTRS_PER_PUD ); }static int __init hugetlb_nrpages_setup (char *s)
Signature · 1 line
1.8.3.1
PATCH v2 2/2 Add /proc/PID/{smaps, numa_maps} support for DAX
So user could check those interface for more detailed information about how much DAX mappings are currently created.
Here we use vma_is_dax method to find specific page struture with DAX {huge, normal}page mappings, vm_normal_page routine works as before without any impact on the existing logical where _vm_normal_page are called.
---
v2:
* Using pte_devmap to check valid pfn page structure, Pointed out by Dan. thx!
--- fs/proc/task_mmu.c | 24 ++++++++++++++++++------ 1 file changed, 18 insertions(+), 6 deletions(-)
fs/proc/task_mmu.c +18 -6
@@ -531,7 +531,10 @@ static void smaps_pte_entry(pte_t *pte, unsigned long addr,struct page *page =NULL ;if (pte_present (*pte)) { - page =vm_normal_page (vma, addr, *pte); +if (!vma_is_dax (vma)) + page =vm_normal_page (vma, addr, *pte); +else if (pte_devmap (*pte)) + page =pte_page (*pte); mss->rss_pte +=PAGE_SIZE ; }else if (is_swap_pte (*pte)) {swp_entry_t swpent =pte_to_swp_entry (*pte);
@@ -583,7 +586,11 @@ static void smaps_pmd_entry(pmd_t *pmd, unsigned long addr,struct page *page;/* FOLL_DUMP will return -EFAULT on huge zero page */ - page =follow_trans_huge_pmd (vma, addr, pmd,FOLL_DUMP ); +if (!vma_is_dax (vma)) + page =follow_trans_huge_pmd (vma, addr, pmd,FOLL_DUMP ); +else if (pmd_devmap (*pmd)) + page =pmd_page (*pmd); +if (IS_ERR_OR_NULL (page))return ;if (PageAnon (page))
@@ -1770,13 +1777,15 @@ static int gather_pte_stats(pmd_t *pmd, unsigned long addr,spinlock_t *ptl;pte_t *orig_pte;pte_t *pte; +struct page *page;#ifdef CONFIG_TRANSPARENT_HUGEPAGE ptl =pmd_trans_huge_lock (pmd, vma);if (ptl) { -struct page *page; - - page =can_gather_numa_stats_pmd (*pmd, vma, addr); +if (!vma_is_dax (vma)) + page =can_gather_numa_stats_pmd (*pmd, vma, addr); +else if (pmd_devmap (*pmd)) + page =pmd_page (*pmd);if (page)gather_stats (page, md,pmd_dirty (*pmd),HPAGE_PMD_SIZE /PAGE_SIZE );
@@ -1789,7 +1798,10 @@ static int gather_pte_stats(pmd_t *pmd, unsigned long addr,#endif orig_pte = pte =pte_offset_map_lock (walk->mm, pmd, addr, &ptl); do { - struct page *page = can_gather_numa_stats(*pte, vma, addr); + if (!vma_is_dax(vma)) + page = can_gather_numa_stats(*pte, vma, addr); + else if (pte_devmap(*pte)) + page = pte_page(*pte); if (!page) continue; gather_stats(page, md, pte_dirty(*pte), 1);
Signature · 1 line
1.8.3.1
kbuild test robot
build botRe: [PATCHv2 2/2] Add /proc/PID/{smaps, numa_maps} support for DAX
Thank you for the patch! Yet we hit a small issue.
Hi Fan,
Thank you for the patch! Yet we hit a small issue.
[auto build test WARNING on linus/master]
[also build test WARNING on v4.14-rc6 next-20171018]
[if your patch is applied to the wrong git tree, please drop us a note to help improve the system]
url: https://github.com/0day-ci/linux/commits/Fan-Du/proc-mm-export-PTE-sizes-directly-in-smaps/20171027-141157
config: x86_64-randconfig-x011-201743 (attached as .config)
compiler: gcc-6 (Debian 6.2.0-3) 6.2.0 20160901 reproduce:
# save the attached .config to linux build tree
make ARCH=x86_64 Note: it may well be a FALSE warning. FWIW you are at least aware of it now.
http://gcc.gnu.org/wiki/Better_Uninitialized_Warnings
All warnings (new ones prefixed by >>):
fs/proc/task_mmu.c: In function 'gather_pte_stats':
>> fs/proc/task_mmu.c:1805:6: warning: 'page' may be used uninitialized in this function [-Wmaybe-uninitialized]
if (!page)
^vim +/page +1805 fs/proc/task_mmu.c
28093f9f3 Gerald Schaefer 2016-04-28 1771
f69ff943d Stephen Wilson 2011-05-24 1772 static int gather_pte_stats(pmd_t *pmd, unsigned long addr,
f69ff943d Stephen Wilson 2011-05-24 1773 unsigned long end, struct mm_walk *walk)
f69ff943d Stephen Wilson 2011-05-24 1774 {
d85f4d6d3 Naoya Horiguchi 2015-02-11 1775 struct numa_maps *md = walk->private;
d85f4d6d3 Naoya Horiguchi 2015-02-11 1776 struct vm_area_struct *vma = walk->vma;
f69ff943d Stephen Wilson 2011-05-24 1777 spinlock_t *ptl;
f69ff943d Stephen Wilson 2011-05-24 1778 pte_t *orig_pte;
f69ff943d Stephen Wilson 2011-05-24 1779 pte_t *pte;
3bf801467 Fan Du 2017-10-24 1780 struct page *page;
f69ff943d Stephen Wilson 2011-05-24 1781
28093f9f3 Gerald Schaefer 2016-04-28 1782 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
b6ec57f4b Kirill A. Shutemov 2016-01-21 1783 ptl = pmd_trans_huge_lock(pmd, vma);
b6ec57f4b Kirill A. Shutemov 2016-01-21 1784 if (ptl) {
3bf801467 Fan Du 2017-10-24 1785 if (!vma_is_dax(vma))
28093f9f3 Gerald Schaefer 2016-04-28 1786 page = can_gather_numa_stats_pmd(*pmd, vma, addr);
3bf801467 Fan Du 2017-10-24 1787 else if (pmd_devmap(*pmd))
3bf801467 Fan Du 2017-10-24 1788 page = pmd_page(*pmd);
32ef43848 Dave Hansen 2011-09-20 1789 if (page)
28093f9f3 Gerald Schaefer 2016-04-28 1790 gather_stats(page, md, pmd_dirty(*pmd),
32ef43848 Dave Hansen 2011-09-20 1791 HPAGE_PMD_SIZE/PAGE_SIZE);
bf929152e Kirill A. Shutemov 2013-11-14 1792 spin_unlock(ptl);
32ef43848 Dave Hansen 2011-09-20 1793 return 0;
32ef43848 Dave Hansen 2011-09-20 1794 }
32ef43848 Dave Hansen 2011-09-20 1795
1a5a9906d Andrea Arcangeli 2012-03-21 1796 if (pmd_trans_unstable(pmd))
1a5a9906d Andrea Arcangeli 2012-03-21 1797 return 0;
28093f9f3 Gerald Schaefer 2016-04-28 1798 #endif
f69ff943d Stephen Wilson 2011-05-24 1799 orig_pte = pte = pte_offset_map_lock(walk->mm, pmd, addr, &ptl);
f69ff943d Stephen Wilson 2011-05-24 1800 do {
3bf801467 Fan Du 2017-10-24 1801 if (!vma_is_dax(vma))
3bf801467 Fan Du 2017-10-24 1802 page = can_gather_numa_stats(*pte, vma, addr);
3bf801467 Fan Du 2017-10-24 1803 else if (pte_devmap(*pte))
3bf801467 Fan Du 2017-10-24 1804 page = pte_page(*pte);
f69ff943d Stephen Wilson 2011-05-24 @1805 if (!page)
f69ff943d Stephen Wilson 2011-05-24 1806 continue;
eb4866d00 Dave Hansen 2011-09-20 1807 gather_stats(page, md, pte_dirty(*pte), 1);
f69ff943d Stephen Wilson 2011-05-24 1808
f69ff943d Stephen Wilson 2011-05-24 1809 } while (pte++, addr += PAGE_SIZE, addr != end);
f69ff943d Stephen Wilson 2011-05-24 1810 pte_unmap_unlock(orig_pte, ptl);
a66c0410b Hugh Dickins 2016-12-12 1811 cond_resched();
f69ff943d Stephen Wilson 2011-05-24 1812 return 0;
f69ff943d Stephen Wilson 2011-05-24 1813 }
f69ff943d Stephen Wilson 2011-05-24 1814 #ifdef CONFIG_HUGETLB_PAGE
632fd60fe Naoya Horiguchi 2015-02-11 1815 static int gather_hugetlb_stats(pte_t *pte, unsigned long hmask,
f69ff943d Stephen Wilson 2011-05-24 1816 unsigned long addr, unsigned long end, struct mm_walk *walk)
f69ff943d Stephen Wilson 2011-05-24 1817 {
5c2ff95e4 Michael Holzheu 2016-02-02 1818 pte_t huge_pte = huge_ptep_get(pte);
f69ff943d Stephen Wilson 2011-05-24 1819 struct numa_maps *md;
f69ff943d Stephen Wilson 2011-05-24 1820 struct page *page;
f69ff943d Stephen Wilson 2011-05-24 1821
5c2ff95e4 Michael Holzheu 2016-02-02 1822 if (!pte_present(huge_pte))
f69ff943d Stephen Wilson 2011-05-24 1823 return 0;
f69ff943d Stephen Wilson 2011-05-24 1824
5c2ff95e4 Michael Holzheu 2016-02-02 1825 page = pte_page(huge_pte);
f69ff943d Stephen Wilson 2011-05-24 1826 if (!page)
f69ff943d Stephen Wilson 2011-05-24 1827 return 0;
f69ff943d Stephen Wilson 2011-05-24 1828
f69ff943d Stephen Wilson 2011-05-24 1829 md = walk->private;
5c2ff95e4 Michael Holzheu 2016-02-02 1830 gather_stats(page, md, pte_dirty(huge_pte), 1);
f69ff943d Stephen Wilson 2011-05-24 1831 return 0;
f69ff943d Stephen Wilson 2011-05-24 1832 }
f69ff943d Stephen Wilson 2011-05-24 1833 :::::: The code at line 1805 was first introduced by commit
:::::: f69ff943df6972aae96c10733b6847fa094d8a59 mm: proc: move show_numa_map() to fs/proc/task_mmu.c
:::::: TO: Stephen Wilson <[email protected]> :::::: CC: Linus Torvalds <[email protected]>
---
0-DAY kernel test infrastructure Open Source Technology Center
https://lists.01.org/pipermail/kbuild-all Intel Corporation