The quilt patch titled
Subject: selftests/vm: cow: add COW tests for collapsing of PTE-mapped anon THP
has been removed from the -mm tree. Its filename was
selftests-vm-cow-add-cow-tests-for-collapsing-of-pte-mapped-anon-thp.patch
This patch was dropped because it was merged into the mm-stable branch
of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm
------------------------------------------------------
From: David Hildenbrand <david@xxxxxxxxxx>
Subject: selftests/vm: cow: add COW tests for collapsing of PTE-mapped anon THP
Date: Wed, 4 Jan 2023 15:49:05 +0100
Currently, anonymous PTE-mapped THPs cannot be collapsed in-place:
collapsing (e.g., via MADV_COLLAPSE) implies allocating a fresh THP and
mapping that new THP via a PMD: as it's a fresh anon THP, it will get the
exclusive flag set on the head page and everybody is happy.
However, if the kernel would ever support in-place collapse of anonymous
THPs (replacing a page table mapping each sub-page of a THP via PTEs with
a single PMD mapping the complete THP), exclusivity information stored for
each sub-page would have to be collapsed accordingly:
(1) All PTEs map !exclusive anon sub-pages: the in-place collapsed THP
must not not have the exclusive flag set on the head page mapped by
the PMD. This is the easiest case to handle ("simply don't set any
exclusive flags").
(2) All PTEs map exclusive anon sub-pages: when collapsing, we have to
clear the exclusive flag from all tail pages and only leave the
exclusive flag set for the head page. Otherwise, fork() after
collapse would not clear the exclusive flags from the tail pages
and we'd be in trouble once PTE-mapping the shared THP when writing
to shared tail pages that still have the exclusive flag set. This
would effectively revert what the PTE-mapping code does when
propagating the exclusive flag to all sub-pages.
(3) PTEs map a mixture of exclusive and !exclusive anon sub-pages (can
happen e.g., due to MADV_DONTFORK before fork()). We must not
collapse the THP in-place, otherwise bad things may happen:
the exclusive flags of sub-pages would get ignored and the
exclusive flag of the head page would get used instead.
Now that we have MADV_COLLAPSE in place to trigger collapsing a THP, let's
add some test cases that would bail out early, if we'd
voluntarily/accidantially unlock in-place collapse for anon THPs and
forget about taking proper care of exclusive flags.
Running the test on a kernel with MADV_COLLAPSE support:
# [INFO] Anonymous THP tests
# [RUN] Basic COW after fork() when collapsing before fork()
ok 169 No leak from parent into child
# [RUN] Basic COW after fork() when collapsing after fork() (fully shared)
ok 170 # SKIP MADV_COLLAPSE failed: Invalid argument
# [RUN] Basic COW after fork() when collapsing after fork() (lower shared)
ok 171 No leak from parent into child
# [RUN] Basic COW after fork() when collapsing after fork() (upper shared)
ok 172 No leak from parent into child
For now, MADV_COLLAPSE always seems to fail if all PTEs map shared
sub-pages.
Link: https://lkml.kernel.org/r/20230104144905.460075-1-david@xxxxxxxxxx
Signed-off-by: David Hildenbrand <david@xxxxxxxxxx>
Cc: Shuah Khan <shuah@xxxxxxxxxx>
Cc: Hugh Dickins <hughd@xxxxxxxxxx>
Cc: Peter Xu <peterx@xxxxxxxxxx>
Cc: Vlastimil Babka <vbabka@xxxxxxx>
Cc: Nadav Amit <nadav.amit@xxxxxxxxx>
Cc: Zach O'Keefe <zokeefe@xxxxxxxxxx>
Cc: Andrea Arcangeli <aarcange@xxxxxxxxxx>
Signed-off-by: Andrew Morton <akpm@xxxxxxxxxxxxxxxxxxxx>
---
tools/testing/selftests/vm/cow.c | 228 +++++++++++++++++++++++++++++
1 file changed, 228 insertions(+)
--- a/tools/testing/selftests/vm/cow.c~selftests-vm-cow-add-cow-tests-for-collapsing-of-pte-mapped-anon-thp
+++ a/tools/testing/selftests/vm/cow.c
@@ -30,6 +30,10 @@
#include "../kselftest.h"
#include "vm_util.h"
+#ifndef MADV_COLLAPSE
+#define MADV_COLLAPSE 25
+#endif
+
static size_t pagesize;
static int pagemap_fd;
static size_t thpsize;
@@ -1178,6 +1182,228 @@ static int tests_per_anon_test_case(void
return tests;
}
+enum anon_thp_collapse_test {
+ ANON_THP_COLLAPSE_UNSHARED,
+ ANON_THP_COLLAPSE_FULLY_SHARED,
+ ANON_THP_COLLAPSE_LOWER_SHARED,
+ ANON_THP_COLLAPSE_UPPER_SHARED,
+};
+
+static void do_test_anon_thp_collapse(char *mem, size_t size,
+ enum anon_thp_collapse_test test)
+{
+ struct comm_pipes comm_pipes;
+ char buf;
+ int ret;
+
+ ret = setup_comm_pipes(&comm_pipes);
+ if (ret) {
+ ksft_test_result_fail("pipe() failed\n");
+ return;
+ }
+
+ /*
+ * Trigger PTE-mapping the THP by temporarily mapping a single subpage
+ * R/O, such that we can try collapsing it later.
+ */
+ ret = mprotect(mem + pagesize, pagesize, PROT_READ);
+ if (ret) {
+ ksft_test_result_fail("mprotect() failed\n");
+ goto close_comm_pipes;
+ }
+ ret = mprotect(mem + pagesize, pagesize, PROT_READ | PROT_WRITE);
+ if (ret) {
+ ksft_test_result_fail("mprotect() failed\n");
+ goto close_comm_pipes;
+ }
+
+ switch (test) {
+ case ANON_THP_COLLAPSE_UNSHARED:
+ /* Collapse before actually COW-sharing the page. */
+ ret = madvise(mem, size, MADV_COLLAPSE);
+ if (ret) {
+ ksft_test_result_skip("MADV_COLLAPSE failed: %s\n",
+ strerror(errno));
+ goto close_comm_pipes;
+ }
+ break;
+ case ANON_THP_COLLAPSE_FULLY_SHARED:
+ /* COW-share the full PTE-mapped THP. */
+ break;
+ case ANON_THP_COLLAPSE_LOWER_SHARED:
+ /* Don't COW-share the upper part of the THP. */
+ ret = madvise(mem + size / 2, size / 2, MADV_DONTFORK);
+ if (ret) {
+ ksft_test_result_fail("MADV_DONTFORK failed\n");
+ goto close_comm_pipes;
+ }
+ break;
+ case ANON_THP_COLLAPSE_UPPER_SHARED:
+ /* Don't COW-share the lower part of the THP. */
+ ret = madvise(mem, size / 2, MADV_DONTFORK);
+ if (ret) {
+ ksft_test_result_fail("MADV_DONTFORK failed\n");
+ goto close_comm_pipes;
+ }
+ break;
+ default:
+ assert(false);
+ }
+
+ ret = fork();
+ if (ret < 0) {
+ ksft_test_result_fail("fork() failed\n");
+ goto close_comm_pipes;
+ } else if (!ret) {
+ switch (test) {
+ case ANON_THP_COLLAPSE_UNSHARED:
+ case ANON_THP_COLLAPSE_FULLY_SHARED:
+ exit(child_memcmp_fn(mem, size, &comm_pipes));
+ break;
+ case ANON_THP_COLLAPSE_LOWER_SHARED:
+ exit(child_memcmp_fn(mem, size / 2, &comm_pipes));
+ break;
+ case ANON_THP_COLLAPSE_UPPER_SHARED:
+ exit(child_memcmp_fn(mem + size / 2, size / 2,
+ &comm_pipes));
+ break;
+ default:
+ assert(false);
+ }
+ }
+
+ while (read(comm_pipes.child_ready[0], &buf, 1) != 1)
+ ;
+
+ switch (test) {
+ case ANON_THP_COLLAPSE_UNSHARED:
+ break;
+ case ANON_THP_COLLAPSE_UPPER_SHARED:
+ case ANON_THP_COLLAPSE_LOWER_SHARED:
+ /*
+ * Revert MADV_DONTFORK such that we merge the VMAs and are
+ * able to actually collapse.
+ */
+ ret = madvise(mem, size, MADV_DOFORK);
+ if (ret) {
+ ksft_test_result_fail("MADV_DOFORK failed\n");
+ write(comm_pipes.parent_ready[1], "0", 1);
+ wait(&ret);
+ goto close_comm_pipes;
+ }
+ /* FALLTHROUGH */
+ case ANON_THP_COLLAPSE_FULLY_SHARED:
+ /* Collapse before anyone modified the COW-shared page. */
+ ret = madvise(mem, size, MADV_COLLAPSE);
+ if (ret) {
+ ksft_test_result_skip("MADV_COLLAPSE failed: %s\n",
+ strerror(errno));
+ write(comm_pipes.parent_ready[1], "0", 1);
+ wait(&ret);
+ goto close_comm_pipes;
+ }
+ break;
+ default:
+ assert(false);
+ }
+
+ /* Modify the page. */
+ memset(mem, 0xff, size);
+ write(comm_pipes.parent_ready[1], "0", 1);
+
+ wait(&ret);
+ if (WIFEXITED(ret))
+ ret = WEXITSTATUS(ret);
+ else
+ ret = -EINVAL;
+
+ ksft_test_result(!ret, "No leak from parent into child\n");
+close_comm_pipes:
+ close_comm_pipes(&comm_pipes);
+}
+
+static void test_anon_thp_collapse_unshared(char *mem, size_t size)
+{
+ do_test_anon_thp_collapse(mem, size, ANON_THP_COLLAPSE_UNSHARED);
+}
+
+static void test_anon_thp_collapse_fully_shared(char *mem, size_t size)
+{
+ do_test_anon_thp_collapse(mem, size, ANON_THP_COLLAPSE_FULLY_SHARED);
+}
+
+static void test_anon_thp_collapse_lower_shared(char *mem, size_t size)
+{
+ do_test_anon_thp_collapse(mem, size, ANON_THP_COLLAPSE_LOWER_SHARED);
+}
+
+static void test_anon_thp_collapse_upper_shared(char *mem, size_t size)
+{
+ do_test_anon_thp_collapse(mem, size, ANON_THP_COLLAPSE_UPPER_SHARED);
+}
+
+/*
+ * Test cases that are specific to anonymous THP: pages in private mappings
+ * that may get shared via COW during fork().
+ */
+static const struct test_case anon_thp_test_cases[] = {
+ /*
+ * Basic COW test for fork() without any GUP when collapsing a THP
+ * before fork().
+ *
+ * Re-mapping a PTE-mapped anon THP using a single PMD ("in-place
+ * collapse") might easily get COW handling wrong when not collapsing
+ * exclusivity information properly.
+ */
+ {
+ "Basic COW after fork() when collapsing before fork()",
+ test_anon_thp_collapse_unshared,
+ },
+ /* Basic COW test, but collapse after COW-sharing a full THP. */
+ {
+ "Basic COW after fork() when collapsing after fork() (fully shared)",
+ test_anon_thp_collapse_fully_shared,
+ },
+ /*
+ * Basic COW test, but collapse after COW-sharing the lower half of a
+ * THP.
+ */
+ {
+ "Basic COW after fork() when collapsing after fork() (lower shared)",
+ test_anon_thp_collapse_lower_shared,
+ },
+ /*
+ * Basic COW test, but collapse after COW-sharing the upper half of a
+ * THP.
+ */
+ {
+ "Basic COW after fork() when collapsing after fork() (upper shared)",
+ test_anon_thp_collapse_upper_shared,
+ },
+};
+
+static void run_anon_thp_test_cases(void)
+{
+ int i;
+
+ if (!thpsize)
+ return;
+
+ ksft_print_msg("[INFO] Anonymous THP tests\n");
+
+ for (i = 0; i < ARRAY_SIZE(anon_thp_test_cases); i++) {
+ struct test_case const *test_case = &anon_thp_test_cases[i];
+
+ ksft_print_msg("[RUN] %s\n", test_case->desc);
+ do_run_with_thp(test_case->fn, THP_RUN_PMD);
+ }
+}
+
+static int tests_per_anon_thp_test_case(void)
+{
+ return thpsize ? 1 : 0;
+}
+
typedef void (*non_anon_test_fn)(char *mem, const char *smem, size_t size);
static void test_cow(char *mem, const char *smem, size_t size)
@@ -1518,6 +1744,7 @@ int main(int argc, char **argv)
ksft_print_header();
ksft_set_plan(ARRAY_SIZE(anon_test_cases) * tests_per_anon_test_case() +
+ ARRAY_SIZE(anon_thp_test_cases) * tests_per_anon_thp_test_case() +
ARRAY_SIZE(non_anon_test_cases) * tests_per_non_anon_test_case());
gup_fd = open("/sys/kernel/debug/gup_test", O_RDWR);
@@ -1526,6 +1753,7 @@ int main(int argc, char **argv)
ksft_exit_fail_msg("opening pagemap failed\n");
run_anon_test_cases();
+ run_anon_thp_test_cases();
run_non_anon_test_cases();
err = ksft_get_fail_cnt();
_
Patches currently in -mm which might be from david@xxxxxxxxxx are
mm-debug_vm_pgtable-more-pte_swp_exclusive-sanity-checks.patch
mm-debug_vm_pgtable-more-pte_swp_exclusive-sanity-checks-fix.patch
alpha-mm-support-__have_arch_pte_swp_exclusive.patch
arc-mm-support-__have_arch_pte_swp_exclusive.patch
arm-mm-support-__have_arch_pte_swp_exclusive.patch
csky-mm-support-__have_arch_pte_swp_exclusive.patch
hexagon-mm-support-__have_arch_pte_swp_exclusive.patch
ia64-mm-support-__have_arch_pte_swp_exclusive.patch
loongarch-mm-support-__have_arch_pte_swp_exclusive.patch
m68k-mm-remove-dummy-__swp-definitions-for-nommu.patch
m68k-mm-support-__have_arch_pte_swp_exclusive.patch
microblaze-mm-support-__have_arch_pte_swp_exclusive.patch
mips-mm-support-__have_arch_pte_swp_exclusive.patch
nios2-mm-refactor-swap-pte-layout.patch
nios2-mm-support-__have_arch_pte_swp_exclusive.patch
openrisc-mm-support-__have_arch_pte_swp_exclusive.patch
parisc-mm-support-__have_arch_pte_swp_exclusive.patch
powerpc-mm-support-__have_arch_pte_swp_exclusive-on-32bit-book3s.patch
powerpc-nohash-mm-support-__have_arch_pte_swp_exclusive.patch
riscv-mm-support-__have_arch_pte_swp_exclusive.patch
sh-mm-support-__have_arch_pte_swp_exclusive.patch
sparc-mm-support-__have_arch_pte_swp_exclusive-on-32bit.patch
sparc-mm-support-__have_arch_pte_swp_exclusive-on-64bit.patch
um-mm-support-__have_arch_pte_swp_exclusive.patch
x86-mm-support-__have_arch_pte_swp_exclusive-also-on-32bit.patch
xtensa-mm-support-__have_arch_pte_swp_exclusive.patch
mm-remove-__have_arch_pte_swp_exclusive.patch
