Re: [PATCH] mm/memory-failure: Disable soft offline for HugeTLB pages by default

[Jiaqi Yan <jiaqiyan@google.com>]

On Wed, Sep 10, 2025 at 9:16 AM Kyle Meyer <kyle.meyer@hpe.com> wrote:
>
> Soft offlining a HugeTLB page reduces the available HugeTLB page pool.
> Since HugeTLB pages are preallocated, reducing the available HugeTLB
> page pool can cause allocation failures.
>
> /proc/sys/vm/enable_soft_offline provides a sysctl interface to
> disable/enable soft offline:
>
> 0 - Soft offline is disabled.
> 1 - Soft offline is enabled.
>
> The current sysctl interface does not distinguish between HugeTLB pages
> and other page types.
>
> Disable soft offline for HugeTLB pages by default (1) and extend the
> sysctl interface to preserve existing behavior (2):
>
> 0 - Soft offline is disabled.
> 1 - Soft offline is enabled (excluding HugeTLB pages).
> 2 - Soft offline is enabled (including HugeTLB pages).
>
> Update documentation for the sysctl interface, reference the sysctl
> interface in the sysfs ABI documentation, and update HugeTLB soft
> offline selftests.
>
> Reported-by: Shawn Fan <shawn.fan@intel.com>
> Suggested-by: Tony Luck <tony.luck@intel.com>
> Signed-off-by: Kyle Meyer <kyle.meyer@hpe.com>
> ---
>
> Tony's original patch disabled soft offline for HugeTLB pages when
> a correctable memory error reported via GHES (with "error threshold
> exceeded" set) happened to be on a HugeTLB page:
>
> https://lore.kernel.org/all/20250904155720.22149-1-tony.luck@intel.com
>
> This patch disables soft offline for HugeTLB pages by default
> (not just from GHES).
>
> ---
>  .../ABI/testing/sysfs-memory-page-offline     |  6 ++++
>  Documentation/admin-guide/sysctl/vm.rst       | 18 ++++++++---
>  mm/memory-failure.c                           | 21 ++++++++++--
>  .../selftests/mm/hugetlb-soft-offline.c       | 32 +++++++++++++------
>  4 files changed, 60 insertions(+), 17 deletions(-)
>
> diff --git a/Documentation/ABI/testing/sysfs-memory-page-offline b/Documentation/ABI/testing/sysfs-memory-page-offline
> index 00f4e35f916f..befb89ae39ec 100644
> --- a/Documentation/ABI/testing/sysfs-memory-page-offline
> +++ b/Documentation/ABI/testing/sysfs-memory-page-offline
> @@ -20,6 +20,12 @@ Description:
>                 number, or a error when the offlining failed.  Reading
>                 the file is not allowed.
>
> +               Soft-offline can be disabled/enabled via sysctl:
> +               /proc/sys/vm/enable_soft_offline
> +
> +               For details, see:
> +               Documentation/admin-guide/sysctl/vm.rst
> +
>  What:          /sys/devices/system/memory/hard_offline_page
>  Date:          Sep 2009
>  KernelVersion: 2.6.33
> diff --git a/Documentation/admin-guide/sysctl/vm.rst b/Documentation/admin-guide/sysctl/vm.rst
> index 4d71211fdad8..ae56372bd604 100644
> --- a/Documentation/admin-guide/sysctl/vm.rst
> +++ b/Documentation/admin-guide/sysctl/vm.rst
> @@ -309,19 +309,29 @@ physical memory) vs performance / capacity implications in transparent and
>  HugeTLB cases.
>
>  For all architectures, enable_soft_offline controls whether to soft offline
> -memory pages.  When set to 1, kernel attempts to soft offline the pages
> -whenever it thinks needed.  When set to 0, kernel returns EOPNOTSUPP to
> -the request to soft offline the pages.  Its default value is 1.
> +memory pages:
> +
> +- 0: Soft offline is disabled.
> +- 1: Soft offline is enabled (excluding HugeTLB pages).
> +- 2: Soft offline is enabled (including HugeTLB pages).

Would it be better to keep/inherit the previous documented behavior "1
- Soft offline is enabled (no matter what type of the page is)"? Thus
it will have no impact to users that are very nervous about corrected
memory errors and willing to lose hugetlb page. Something like:

  enum soft_offline {
      SOFT_OFFLINE_DISABLED = 0,
      SOFT_OFFLINE_ENABLED,
      SOFT_OFFLINE_ENABLED_SKIP_HUGETLB,
      // SOFT_OFFLINE_ENABLED_SKIP_XXX...
  };

> +
> +The default is 1.
> +
> +If soft offline is disabled for the requested page type, EOPNOTSUPP is returned.
>
>  It is worth mentioning that after setting enable_soft_offline to 0, the
>  following requests to soft offline pages will not be performed:
>
> +- Request to soft offline from sysfs (soft_offline_page).
> +
>  - Request to soft offline pages from RAS Correctable Errors Collector.
>
> -- On ARM, the request to soft offline pages from GHES driver.
> +- On ARM and X86, the request to soft offline pages from GHES driver.
>
>  - On PARISC, the request to soft offline pages from Page Deallocation Table.
>
> +Note: Soft offlining a HugeTLB page reduces the HugeTLB page pool.
> +
>  extfrag_threshold
>  =================
>
> diff --git a/mm/memory-failure.c b/mm/memory-failure.c
> index fc30ca4804bf..cb59a99b48c5 100644
> --- a/mm/memory-failure.c
> +++ b/mm/memory-failure.c
> @@ -64,11 +64,18 @@
>  #include "internal.h"
>  #include "ras/ras_event.h"
>
> +enum soft_offline {
> +       SOFT_OFFLINE_DISABLED = 0,
> +       SOFT_OFFLINE_ENABLED_SKIP_HUGETLB,
> +       SOFT_OFFLINE_ENABLED
> +};
> +
>  static int sysctl_memory_failure_early_kill __read_mostly;
>
>  static int sysctl_memory_failure_recovery __read_mostly = 1;
>
> -static int sysctl_enable_soft_offline __read_mostly = 1;
> +static int sysctl_enable_soft_offline __read_mostly =
> +       SOFT_OFFLINE_ENABLED_SKIP_HUGETLB;
>
>  atomic_long_t num_poisoned_pages __read_mostly = ATOMIC_LONG_INIT(0);
>
> @@ -150,7 +157,7 @@ static const struct ctl_table memory_failure_table[] = {
>                 .mode           = 0644,
>                 .proc_handler   = proc_dointvec_minmax,
>                 .extra1         = SYSCTL_ZERO,
> -               .extra2         = SYSCTL_ONE,
> +               .extra2         = SYSCTL_TWO,
>         }
>  };
>
> @@ -2799,12 +2806,20 @@ int soft_offline_page(unsigned long pfn, int flags)
>                 return -EIO;
>         }
>
> -       if (!sysctl_enable_soft_offline) {
> +       if (sysctl_enable_soft_offline == SOFT_OFFLINE_DISABLED) {
>                 pr_info_once("disabled by /proc/sys/vm/enable_soft_offline\n");
>                 put_ref_page(pfn, flags);
>                 return -EOPNOTSUPP;
>         }
>
> +       if (sysctl_enable_soft_offline == SOFT_OFFLINE_ENABLED_SKIP_HUGETLB) {
> +               if (folio_test_hugetlb(pfn_folio(pfn))) {
> +                       pr_info_once("disabled for HugeTLB pages by /proc/sys/vm/enable_soft_offline\n");
> +                       put_ref_page(pfn, flags);
> +                       return -EOPNOTSUPP;
> +               }
> +       }
> +
>         mutex_lock(&mf_mutex);
>
>         if (PageHWPoison(page)) {
> diff --git a/tools/testing/selftests/mm/hugetlb-soft-offline.c b/tools/testing/selftests/mm/hugetlb-soft-offline.c
> index f086f0e04756..7e2873cd0a6d 100644
> --- a/tools/testing/selftests/mm/hugetlb-soft-offline.c
> +++ b/tools/testing/selftests/mm/hugetlb-soft-offline.c
> @@ -1,10 +1,15 @@
>  // SPDX-License-Identifier: GPL-2.0
>  /*
>   * Test soft offline behavior for HugeTLB pages:
> - * - if enable_soft_offline = 0, hugepages should stay intact and soft
> - *   offlining failed with EOPNOTSUPP.
> - * - if enable_soft_offline = 1, a hugepage should be dissolved and
> - *   nr_hugepages/free_hugepages should be reduced by 1.
> + *
> + * - if enable_soft_offline = 0 (SOFT_OFFLINE_DISABLED), HugeTLB pages
> + *   should stay intact and soft offlining failed with EOPNOTSUPP.
> + *
> + * - if enable_soft_offline = 1 (SOFT_OFFLINE_ENABLED_SKIP_HUGETLB), HugeTLB pages
> + *   should stay intact and soft offlining failed with EOPNOTSUPP.
> + *
> + * - if enable_soft_offline = 2 (SOFT_OFFLINE_ENABLED), a HugeTLB page should be
> + *   dissolved and nr_hugepages/free_hugepages should be reduced by 1.
>   *
>   * Before running, make sure more than 2 hugepages of default_hugepagesz
>   * are allocated. For example, if /proc/meminfo/Hugepagesize is 2048kB:
> @@ -32,6 +37,12 @@
>
>  #define EPREFIX " !!! "
>
> +enum soft_offline {
> +       SOFT_OFFLINE_DISABLED = 0,
> +       SOFT_OFFLINE_ENABLED_SKIP_HUGETLB,
> +       SOFT_OFFLINE_ENABLED
> +};
> +
>  static int do_soft_offline(int fd, size_t len, int expect_errno)
>  {
>         char *filemap = NULL;
> @@ -83,7 +94,7 @@ static int set_enable_soft_offline(int value)
>         char cmd[256] = {0};
>         FILE *cmdfile = NULL;
>
> -       if (value != 0 && value != 1)
> +       if (value < SOFT_OFFLINE_DISABLED || value > SOFT_OFFLINE_ENABLED)
>                 return -EINVAL;
>
>         sprintf(cmd, "echo %d > /proc/sys/vm/enable_soft_offline", value);
> @@ -155,7 +166,7 @@ static int create_hugetlbfs_file(struct statfs *file_stat)
>  static void test_soft_offline_common(int enable_soft_offline)
>  {
>         int fd;
> -       int expect_errno = enable_soft_offline ? 0 : EOPNOTSUPP;
> +       int expect_errno = (enable_soft_offline == SOFT_OFFLINE_ENABLED) ? 0 : EOPNOTSUPP;
>         struct statfs file_stat;
>         unsigned long hugepagesize_kb = 0;
>         unsigned long nr_hugepages_before = 0;
> @@ -198,7 +209,7 @@ static void test_soft_offline_common(int enable_soft_offline)
>         // No need for the hugetlbfs file from now on.
>         close(fd);
>
> -       if (enable_soft_offline) {
> +       if (enable_soft_offline == SOFT_OFFLINE_ENABLED) {
>                 if (nr_hugepages_before != nr_hugepages_after + 1) {
>                         ksft_test_result_fail("MADV_SOFT_OFFLINE should reduced 1 hugepage\n");
>                         return;
> @@ -219,10 +230,11 @@ static void test_soft_offline_common(int enable_soft_offline)
>  int main(int argc, char **argv)
>  {
>         ksft_print_header();
> -       ksft_set_plan(2);
> +       ksft_set_plan(3);
>
> -       test_soft_offline_common(1);
> -       test_soft_offline_common(0);
> +       test_soft_offline_common(SOFT_OFFLINE_ENABLED);
> +       test_soft_offline_common(SOFT_OFFLINE_ENABLED_SKIP_HUGETLB);
> +       test_soft_offline_common(SOFT_OFFLINE_DISABLED);

Thanks for updating the test code! Looks good to me.

>
>         ksft_finished();
>  }
> --
> 2.51.0
>