diff options
author | David Hildenbrand <david@redhat.com> | 2024-06-05 11:17:10 +0200 |
---|---|---|
committer | Andrew Morton <akpm@linux-foundation.org> | 2024-06-15 10:43:05 -0700 |
commit | 384a746bb55960aa5ffb3a67de08f11fc2f51042 (patch) | |
tree | 9ccb04eb8cff6811cfd609be1deb332082e158f1 | |
parent | 8bb592c2eca8fd2bc06db7d80b38da18da4a2f43 (diff) |
Revert "mm: init_mlocked_on_free_v3"
There was insufficient review and no agreement that this is the right
approach.
There are serious flaws with the implementation that make processes using
mlock() not even work with simple fork() [1] and we get reliable crashes
when rebooting.
Further, simply because we might be unmapping a single PTE of a large
mlocked folio, we shouldn't zero out the whole folio.
... especially because the code can also *corrupt* urelated memory because
kernel_init_pages(page, folio_nr_pages(folio));
Could end up writing outside of the actual folio if we work with a tail
page.
Let's revert it. Once there is agreement that this is the right approach,
the issues were fixed and there was reasonable review and proper testing,
we can consider it again.
[1] https://lkml.kernel.org/r/4da9da2f-73e4-45fd-b62f-a8a513314057@redhat.com
Link: https://lkml.kernel.org/r/20240605091710.38961-1-david@redhat.com
Fixes: ba42b524a040 ("mm: init_mlocked_on_free_v3")
Signed-off-by: David Hildenbrand <david@redhat.com>
Reported-by: David Wang <00107082@163.com>
Closes: https://lore.kernel.org/lkml/20240528151340.4282-1-00107082@163.com/
Reported-by: Lance Yang <ioworker0@gmail.com>
Closes: https://lkml.kernel.org/r/20240601140917.43562-1-ioworker0@gmail.com
Acked-by: Lance Yang <ioworker0@gmail.com>
Cc: York Jasper Niebuhr <yjnworkstation@gmail.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Kees Cook <keescook@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
-rw-r--r-- | Documentation/admin-guide/kernel-parameters.txt | 6 | ||||
-rw-r--r-- | include/linux/mm.h | 9 | ||||
-rw-r--r-- | mm/internal.h | 1 | ||||
-rw-r--r-- | mm/memory.c | 6 | ||||
-rw-r--r-- | mm/mm_init.c | 43 | ||||
-rw-r--r-- | mm/page_alloc.c | 2 | ||||
-rw-r--r-- | security/Kconfig.hardening | 15 |
7 files changed, 9 insertions, 73 deletions
diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt index b600df82669d..11e57ba2985c 100644 --- a/Documentation/admin-guide/kernel-parameters.txt +++ b/Documentation/admin-guide/kernel-parameters.txt @@ -2192,12 +2192,6 @@ Format: 0 | 1 Default set by CONFIG_INIT_ON_FREE_DEFAULT_ON. - init_mlocked_on_free= [MM] Fill freed userspace memory with zeroes if - it was mlock'ed and not explicitly munlock'ed - afterwards. - Format: 0 | 1 - Default set by CONFIG_INIT_MLOCKED_ON_FREE_DEFAULT_ON - init_pkru= [X86] Specify the default memory protection keys rights register contents for all processes. 0x55555554 by default (disallow access to all but pkey 0). Can diff --git a/include/linux/mm.h b/include/linux/mm.h index 9849dfda44d4..9a5652c5fadd 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -3776,14 +3776,7 @@ DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free); static inline bool want_init_on_free(void) { return static_branch_maybe(CONFIG_INIT_ON_FREE_DEFAULT_ON, - &init_on_free); -} - -DECLARE_STATIC_KEY_MAYBE(CONFIG_INIT_MLOCKED_ON_FREE_DEFAULT_ON, init_mlocked_on_free); -static inline bool want_init_mlocked_on_free(void) -{ - return static_branch_maybe(CONFIG_INIT_MLOCKED_ON_FREE_DEFAULT_ON, - &init_mlocked_on_free); + &init_on_free); } extern bool _debug_pagealloc_enabled_early; diff --git a/mm/internal.h b/mm/internal.h index b2c75b12014e..c72c306761a4 100644 --- a/mm/internal.h +++ b/mm/internal.h @@ -588,7 +588,6 @@ extern void __putback_isolated_page(struct page *page, unsigned int order, extern void memblock_free_pages(struct page *page, unsigned long pfn, unsigned int order); extern void __free_pages_core(struct page *page, unsigned int order); -extern void kernel_init_pages(struct page *page, int numpages); /* * This will have no effect, other than possibly generating a warning, if the diff --git a/mm/memory.c b/mm/memory.c index 0f47a533014e..2bc8032a30a2 100644 --- a/mm/memory.c +++ b/mm/memory.c @@ -1507,12 +1507,6 @@ static __always_inline void zap_present_folio_ptes(struct mmu_gather *tlb, if (unlikely(folio_mapcount(folio) < 0)) print_bad_pte(vma, addr, ptent, page); } - - if (want_init_mlocked_on_free() && folio_test_mlocked(folio) && - !delay_rmap && folio_test_anon(folio)) { - kernel_init_pages(page, folio_nr_pages(folio)); - } - if (unlikely(__tlb_remove_folio_pages(tlb, page, nr, delay_rmap))) { *force_flush = true; *force_break = true; diff --git a/mm/mm_init.c b/mm/mm_init.c index f72b852bd5b8..3ec04933f7fd 100644 --- a/mm/mm_init.c +++ b/mm/mm_init.c @@ -2523,9 +2523,6 @@ EXPORT_SYMBOL(init_on_alloc); DEFINE_STATIC_KEY_MAYBE(CONFIG_INIT_ON_FREE_DEFAULT_ON, init_on_free); EXPORT_SYMBOL(init_on_free); -DEFINE_STATIC_KEY_MAYBE(CONFIG_INIT_MLOCKED_ON_FREE_DEFAULT_ON, init_mlocked_on_free); -EXPORT_SYMBOL(init_mlocked_on_free); - static bool _init_on_alloc_enabled_early __read_mostly = IS_ENABLED(CONFIG_INIT_ON_ALLOC_DEFAULT_ON); static int __init early_init_on_alloc(char *buf) @@ -2543,14 +2540,6 @@ static int __init early_init_on_free(char *buf) } early_param("init_on_free", early_init_on_free); -static bool _init_mlocked_on_free_enabled_early __read_mostly - = IS_ENABLED(CONFIG_INIT_MLOCKED_ON_FREE_DEFAULT_ON); -static int __init early_init_mlocked_on_free(char *buf) -{ - return kstrtobool(buf, &_init_mlocked_on_free_enabled_early); -} -early_param("init_mlocked_on_free", early_init_mlocked_on_free); - DEFINE_STATIC_KEY_MAYBE(CONFIG_DEBUG_VM, check_pages_enabled); /* @@ -2578,21 +2567,12 @@ static void __init mem_debugging_and_hardening_init(void) } #endif - if ((_init_on_alloc_enabled_early || _init_on_free_enabled_early || - _init_mlocked_on_free_enabled_early) && + if ((_init_on_alloc_enabled_early || _init_on_free_enabled_early) && page_poisoning_requested) { pr_info("mem auto-init: CONFIG_PAGE_POISONING is on, " - "will take precedence over init_on_alloc, init_on_free " - "and init_mlocked_on_free\n"); + "will take precedence over init_on_alloc and init_on_free\n"); _init_on_alloc_enabled_early = false; _init_on_free_enabled_early = false; - _init_mlocked_on_free_enabled_early = false; - } - - if (_init_mlocked_on_free_enabled_early && _init_on_free_enabled_early) { - pr_info("mem auto-init: init_on_free is on, " - "will take precedence over init_mlocked_on_free\n"); - _init_mlocked_on_free_enabled_early = false; } if (_init_on_alloc_enabled_early) { @@ -2609,17 +2589,9 @@ static void __init mem_debugging_and_hardening_init(void) static_branch_disable(&init_on_free); } - if (_init_mlocked_on_free_enabled_early) { - want_check_pages = true; - static_branch_enable(&init_mlocked_on_free); - } else { - static_branch_disable(&init_mlocked_on_free); - } - - if (IS_ENABLED(CONFIG_KMSAN) && (_init_on_alloc_enabled_early || - _init_on_free_enabled_early || _init_mlocked_on_free_enabled_early)) - pr_info("mem auto-init: please make sure init_on_alloc, init_on_free and " - "init_mlocked_on_free are disabled when running KMSAN\n"); + if (IS_ENABLED(CONFIG_KMSAN) && + (_init_on_alloc_enabled_early || _init_on_free_enabled_early)) + pr_info("mem auto-init: please make sure init_on_alloc and init_on_free are disabled when running KMSAN\n"); #ifdef CONFIG_DEBUG_PAGEALLOC if (debug_pagealloc_enabled()) { @@ -2658,10 +2630,9 @@ static void __init report_meminit(void) else stack = "off"; - pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s, mlocked free:%s\n", + pr_info("mem auto-init: stack:%s, heap alloc:%s, heap free:%s\n", stack, want_init_on_alloc(GFP_KERNEL) ? "on" : "off", - want_init_on_free() ? "on" : "off", - want_init_mlocked_on_free() ? "on" : "off"); + want_init_on_free() ? "on" : "off"); if (want_init_on_free()) pr_info("mem auto-init: clearing system memory may take some time...\n"); } diff --git a/mm/page_alloc.c b/mm/page_alloc.c index 222299b5c0e6..7300aa9f14b0 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c @@ -1016,7 +1016,7 @@ static inline bool should_skip_kasan_poison(struct page *page) return page_kasan_tag(page) == KASAN_TAG_KERNEL; } -void kernel_init_pages(struct page *page, int numpages) +static void kernel_init_pages(struct page *page, int numpages) { int i; diff --git a/security/Kconfig.hardening b/security/Kconfig.hardening index effbf5982be1..2cff851ebfd7 100644 --- a/security/Kconfig.hardening +++ b/security/Kconfig.hardening @@ -255,21 +255,6 @@ config INIT_ON_FREE_DEFAULT_ON touching "cold" memory areas. Most cases see 3-5% impact. Some synthetic workloads have measured as high as 8%. -config INIT_MLOCKED_ON_FREE_DEFAULT_ON - bool "Enable mlocked memory zeroing on free" - depends on !KMSAN - help - This config has the effect of setting "init_mlocked_on_free=1" - on the kernel command line. If it is enabled, all mlocked process - memory is zeroed when freed. This restriction to mlocked memory - improves performance over "init_on_free" but can still be used to - protect confidential data like key material from content exposures - to other processes, as well as live forensics and cold boot attacks. - Any non-mlocked memory is not cleared before it is reassigned. This - configuration can be overwritten by setting "init_mlocked_on_free=0" - on the command line. The "init_on_free" boot option takes - precedence over "init_mlocked_on_free". - config CC_HAS_ZERO_CALL_USED_REGS def_bool $(cc-option,-fzero-call-used-regs=used-gpr) # https://github.com/ClangBuiltLinux/linux/issues/1766 |