mm/khugepaged: fix GUP-fast interaction by sending IPI
authorJann Horn <jannh@google.com>
Fri, 25 Nov 2022 21:37:13 +0000 (22:37 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 7 Jan 2023 11:07:10 +0000 (12:07 +0100)
commit 2ba99c5e08812494bc57f319fb562f527d9bacd8 upstream.

Since commit 70cbc3cc78a99 ("mm: gup: fix the fast GUP race against THP
collapse"), the lockless_pages_from_mm() fastpath rechecks the pmd_t to
ensure that the page table was not removed by khugepaged in between.

However, lockless_pages_from_mm() still requires that the page table is
not concurrently freed.  Fix it by sending IPIs (if the architecture uses
semi-RCU-style page table freeing) before freeing/reusing page tables.

Link: https://lkml.kernel.org/r/20221129154730.2274278-2-jannh@google.com
Link: https://lkml.kernel.org/r/20221128180252.1684965-2-jannh@google.com
Link: https://lkml.kernel.org/r/20221125213714.4115729-2-jannh@google.com
Fixes: ba76149f47d8 ("thp: khugepaged")
Signed-off-by: Jann Horn <jannh@google.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[manual backport: two of the three places in khugepaged that can free
ptes were refactored into a common helper between 5.15 and 6.0;
TLB flushing was refactored between 5.4 and 5.10;
TLB flushing was refactored between 4.19 and 5.4;
pmd collapse for PTE-mapped THP was only added in 5.4;
ugly hack for s390 in <=4.19 and arm]
Signed-off-by: Jann Horn <jannh@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
include/asm-generic/tlb.h
mm/khugepaged.c
mm/memory.c

index e9851100c0f7e0493ae5087d0553bae0ad57fef2..4b1c142be08c5e6f4e9543fd67c01bcf4ac74756 100644 (file)
@@ -60,6 +60,12 @@ struct mmu_table_batch {
 extern void tlb_table_flush(struct mmu_gather *tlb);
 extern void tlb_remove_table(struct mmu_gather *tlb, void *table);
 
+void tlb_remove_table_sync_one(void);
+
+#else
+
+static inline void tlb_remove_table_sync_one(void) { }
+
 #endif
 
 /*
index 0f1bdbae45e21a12dccbf3fb0325b247d7ee3b00..447cd59749414ad79c371b8e1a890eac2e845940 100644 (file)
 #include <asm/pgalloc.h>
 #include "internal.h"
 
+/* gross hack for <=4.19 stable */
+#if defined(CONFIG_S390) || defined(CONFIG_ARM)
+static void tlb_remove_table_smp_sync(void *arg)
+{
+        /* Simply deliver the interrupt */
+}
+
+static void tlb_remove_table_sync_one(void)
+{
+        smp_call_function(tlb_remove_table_smp_sync, NULL, 1);
+}
+#endif
+
 enum scan_result {
        SCAN_FAIL,
        SCAN_SUCCEED,
@@ -1044,6 +1057,7 @@ static void collapse_huge_page(struct mm_struct *mm,
        _pmd = pmdp_collapse_flush(vma, address, pmd);
        spin_unlock(pmd_ptl);
        mmu_notifier_invalidate_range_end(mm, mmun_start, mmun_end);
+       tlb_remove_table_sync_one();
 
        spin_lock(pte_ptl);
        isolated = __collapse_huge_page_isolate(vma, address, pte);
@@ -1293,6 +1307,7 @@ static void retract_page_tables(struct address_space *mapping, pgoff_t pgoff)
                                _pmd = pmdp_collapse_flush(vma, addr, pmd);
                                spin_unlock(ptl);
                                atomic_long_dec(&mm->nr_ptes);
+                               tlb_remove_table_sync_one();
                                pte_free(mm, pmd_pgtable(_pmd));
                        }
                        up_write(&mm->mmap_sem);
index 36d46e19df9607504aa0c35965fbbf8b82388529..a93ea671b8f17400f2121ecc860efbbe93ba8d87 100644 (file)
@@ -349,6 +349,11 @@ static void tlb_remove_table_smp_sync(void *arg)
        /* Simply deliver the interrupt */
 }
 
+void tlb_remove_table_sync_one(void)
+{
+       smp_call_function(tlb_remove_table_smp_sync, NULL, 1);
+}
+
 static void tlb_remove_table_one(void *table)
 {
        /*