Re: [PATCH] ARM64: Implement arch_report_meminfo()

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On 14/12/2023 9:35 pm, Christoph Lameter (Ampere) wrote:
On Thu, 14 Dec 2023, Robin Murphy wrote:

It seems somewhat suspect that these counts only ever increase. It's not often that we change or remove parts of the linear map, but it certainly can happen.

Well yes in the case of hotplug I guess ... Ok here is V2

There are also paths where we remove and reinstate parts of the linear map via set_memory_valid(), set_direct_map_*(), and possibly others. If we're exposing a user ABI that claims to be accounting kernel VA mappings, then I think users are within their rights to expect it to actually account kernel VA mappings, not just expose numbers whose only guaranteed significance is whether they are zero or nonzero.

Looking again, am I also right in thinking that what I assumed were the non-contiguous counts here are actually total counts of *either* type of mapping at that level, and inclusive of the contiguous counts? If so, that seems a bit non-obvious - my intuitive expectation would be for the sum of all these numbers to represent the total amount of direct-mapped RAM, where either we're intersted in each distinct type of mapping and accounting them all separately, or we're simply interested in the general shape of the pagetables, and thus would account per-level and ignore the contiguous bit since we don't know whether it's actually doing anything useful anyweay.

Thanks,
Robin.

 From cl@xxxxxxxxxx Fri Dec  8 13:11:58 2023
From: "Christoph Lameter (Ampere)" <cl@xxxxxxxxxx>
To: Catalin Marinas <catalin.marinas@xxxxxxx>, Marc Zyngier <maz@xxxxxxxxxx>, Will Deacon <will@xxxxxxxxxx>, Ryan Roberts <ryan.roberts@xxxxxxx>, Mark Rutland <mark.rutland@xxxxxxx>, Vishal Moola <vishal.moola@xxxxxxxxx>
Cc: linux-arm-kernel@xxxxxxxxxxxxxxxxxxx, linux-mm@xxxxxxxxx
Subject: [PATCH] ARM64: Implement arch_report_meminfo()

V1->V2 hotplug and umapping support

X86 has information in /proc/meminfo showing the use of large mappings
for the kernel direct map. This has now also become important for
ARM since the kernel default CONFIG_RODATA_FULL_DEFAULT_ENABLED
forces 4K PTE use for the direct map and users may not be aware
of the performance impact of the increased TLB use etc.

The output of /proc/meminfo on ARM64 is then after this patch:

4K page size:

Hugepagesize:       2048 kB
Hugetlb:               0 kB
DirectMap4k:      155912 kB
CONT DMap4k:        1176 kB
DirectMap2M:      722944 kB
CONT DMap2M:       28672 kB
DirectMap1G:   534773760 kB

64K page size:

Hugepagesize:     524288 kB
Hugetlb:               0 kB
DirectMap64k:      882624 kB
CONT DMap64k:       19904 kB
DirectMap512M:    534773760 kB
CONT DMap512M:           0 kB
DirectMap4096G:           0 kB

Signed-off-by: Christoph Lameter (Ampere) <cl@xxxxxxxxx>

Index: linux/arch/arm64/mm/mmu.c
===================================================================
--- linux.orig/arch/arm64/mm/mmu.c
+++ linux/arch/arm64/mm/mmu.c
@@ -76,6 +76,13 @@ EXPORT_SYMBOL(empty_zero_page);
  static DEFINE_SPINLOCK(swapper_pgdir_lock);
  static DEFINE_MUTEX(fixmap_lock);

+static atomic_t nr_pte;
+static atomic_t nr_pmd;
+static atomic_t nr_pud;
+static atomic_t nr_pte_cont;
+static atomic_t nr_pmd_cont;
+
+
  void set_swapper_pgd(pgd_t *pgdp, pgd_t pgd)
  {
      pgd_t *fixmap_pgdp;
@@ -179,6 +186,7 @@ static void init_pte(pmd_t *pmdp, unsign
          pte_t old_pte = READ_ONCE(*ptep);

          set_pte(ptep, pfn_pte(__phys_to_pfn(phys), prot));
+        atomic_inc(&nr_pte);

          /*
           * After the PTE entry has been populated once, we
@@ -223,8 +231,10 @@ static void alloc_init_cont_pte(pmd_t *p

          /* use a contiguous mapping if the range is suitably aligned */
          if ((((addr | next | phys) & ~CONT_PTE_MASK) == 0) &&
-            (flags & NO_CONT_MAPPINGS) == 0)
+            (flags & NO_CONT_MAPPINGS) == 0) {
              __prot = __pgprot(pgprot_val(prot) | PTE_CONT);
+            atomic_inc(&nr_pte_cont);
+        }

          init_pte(pmdp, addr, next, phys, __prot);

@@ -249,6 +259,7 @@ static void init_pmd(pud_t *pudp, unsign
          if (((addr | next | phys) & ~PMD_MASK) == 0 &&
              (flags & NO_BLOCK_MAPPINGS) == 0) {
              pmd_set_huge(pmdp, phys, prot);
+            atomic_inc(&nr_pmd);

              /*
               * After the PMD entry has been populated once, we
@@ -301,8 +312,10 @@ static void alloc_init_cont_pmd(pud_t *p

          /* use a contiguous mapping if the range is suitably aligned */
          if ((((addr | next | phys) & ~CONT_PMD_MASK) == 0) &&
-            (flags & NO_CONT_MAPPINGS) == 0)
+            (flags & NO_CONT_MAPPINGS) == 0) {
              __prot = __pgprot(pgprot_val(prot) | PTE_CONT);
+            atomic_inc(&nr_pmd_cont);
+        }

          init_pmd(pudp, addr, next, phys, __prot, pgtable_alloc, flags);

@@ -346,7 +359,7 @@ static void alloc_init_pud(pgd_t *pgdp,
             ((addr | next | phys) & ~PUD_MASK) == 0 &&
              (flags & NO_BLOCK_MAPPINGS) == 0) {
              pud_set_huge(pudp, phys, prot);
-
+            atomic_inc(&nr_pud);
              /*
               * After the PUD entry has been populated once, we
               * only allow updates to the permission attributes.
@@ -859,6 +872,11 @@ static void unmap_hotplug_pte_range(pmd_
              continue;

          WARN_ON(!pte_present(pte));
+
+        if (pte_cont(pte))
+            atomic_dec(&nr_pte_cont);
+        atomic_dec(&nr_pte);
+
          pte_clear(&init_mm, addr, ptep);
          flush_tlb_kernel_range(addr, addr + PAGE_SIZE);
          if (free_mapped)
@@ -883,6 +901,11 @@ static void unmap_hotplug_pmd_range(pud_

          WARN_ON(!pmd_present(pmd));
          if (pmd_sect(pmd)) {
+
+            if (pmd_cont(pmd))
+                atomic_dec(&nr_pmd_cont);
+            atomic_dec(&nr_pmd);
+
              pmd_clear(pmdp);

              /*
@@ -916,6 +939,9 @@ static void unmap_hotplug_pud_range(p4d_

          WARN_ON(!pud_present(pud));
          if (pud_sect(pud)) {
+
+            atomic_dec(&nr_pud);
+
              pud_clear(pudp);

              /*
@@ -1010,6 +1036,7 @@ static void free_empty_pte_table(pmd_t *
              return;
      }

+    atomic_dec(&nr_pmd);
      pmd_clear(pmdp);
      __flush_tlb_kernel_pgtable(start);
      free_hotplug_pgtable_page(virt_to_page(ptep));
@@ -1050,6 +1077,7 @@ static void free_empty_pmd_table(pud_t *
              return;
      }

+    atomic_dec(&nr_pud);
      pud_clear(pudp);
      __flush_tlb_kernel_pgtable(start);
      free_hotplug_pgtable_page(virt_to_page(pmdp));
@@ -1225,6 +1253,7 @@ int pmd_free_pte_page(pmd_t *pmdp, unsig
      }

      table = pte_offset_kernel(pmdp, addr);
+    atomic_dec(&nr_pmd);
      pmd_clear(pmdp);
      __flush_tlb_kernel_pgtable(addr);
      pte_free_kernel(NULL, table);
@@ -1253,6 +1282,7 @@ int pud_free_pmd_page(pud_t *pudp, unsig
          pmd_free_pte_page(pmdp, next);
      } while (pmdp++, next += PMD_SIZE, next != end);

+    atomic_dec(&nr_pud);
      pud_clear(pudp);
      __flush_tlb_kernel_pgtable(addr);
      pmd_free(NULL, table);
@@ -1486,3 +1516,36 @@ void ptep_modify_prot_commit(struct vm_a
  {
      set_pte_at(vma->vm_mm, addr, ptep, pte);
  }
+
+
+#ifdef CONFIG_PROC_FS
+void arch_report_meminfo(struct seq_file *m)
+{
+    unsigned long pagesize_in_kb = PAGE_SIZE / 1024;
+
+    seq_printf(m, "DirectMap%luk:    %8lu kB\n",
+            pagesize_in_kb,
+              (unsigned long)atomic_read(&nr_pte) * pagesize_in_kb);
+
+        seq_printf(m, "CONT DMap%luk:    %8lu kB\n",
+            pagesize_in_kb,
+            (unsigned long)atomic_read(&nr_pte_cont) * pagesize_in_kb);
+
+    pagesize_in_kb = PMD_SIZE / 1024;
+
+        seq_printf(m, "DirectMap%luM:    %8lu kB\n",
+            pagesize_in_kb / 1024,
+            (unsigned long)atomic_read(&nr_pmd) * pagesize_in_kb);
+
+    seq_printf(m, "CONT DMap%luM:    %8lu kB\n",
+            pagesize_in_kb / 1024,
+            (unsigned long)atomic_read(&nr_pmd_cont) * pagesize_in_kb);
+
+    pagesize_in_kb = PUD_SIZE / 1024;
+
+    seq_printf(m, "DirectMap%luG:  %10lu kB\n",
+            pagesize_in_kb >> 20,
+            (unsigned long)atomic_read(&nr_pud) * pagesize_in_kb);
+}
+#endif
+




[Index of Archives]     [Linux ARM Kernel]     [Linux ARM]     [Linux Omap]     [Fedora ARM]     [IETF Annouce]     [Bugtraq]     [Linux OMAP]     [Linux MIPS]     [eCos]     [Asterisk Internet PBX]     [Linux API]

  Powered by Linux