KAISER comes with overhead. The most expensive part is the CR3 switching in the entry code. Add a command line parameter which allows to disable KAISER at boot time. Most code pathes simply check a variable, but the entry code uses a static branch. The other code pathes cannot use a static branch because they are used before jump label patching is possible. Not an issue as the code pathes are not so performance sensitive as the entry/exit code. This makes KAISER depend on JUMP_LABEL and on a GCC which supports it, but that's a resonable requirement. The PGD allocation is still 8k when CONFIG_KAISER is enabled. This can be addressed on top of this. Signed-off-by: Thomas Gleixner --- arch/x86/entry/calling.h | 7 +++++++ arch/x86/include/asm/kaiser.h | 10 ++++++++++ arch/x86/include/asm/pgtable_64.h | 6 ++++++ arch/x86/mm/dump_pagetables.c | 5 ++++- arch/x86/mm/init.c | 7 ++++--- arch/x86/mm/kaiser.c | 30 ++++++++++++++++++++++++++++++ security/Kconfig | 2 +- 7 files changed, 62 insertions(+), 5 deletions(-) --- a/arch/x86/entry/calling.h +++ b/arch/x86/entry/calling.h @@ -210,18 +210,23 @@ For 32-bit we have the following convent .endm .macro SWITCH_TO_KERNEL_CR3 scratch_reg:req + STATIC_JUMP_IF_FALSE .Lend_\@, kaiser_enabled_key, def=1 mov %cr3, \scratch_reg ADJUST_KERNEL_CR3 \scratch_reg mov \scratch_reg, %cr3 +.Lend_\@: .endm .macro SWITCH_TO_USER_CR3 scratch_reg:req + STATIC_JUMP_IF_FALSE .Lend_\@, kaiser_enabled_key, def=1 mov %cr3, \scratch_reg ADJUST_USER_CR3 \scratch_reg mov \scratch_reg, %cr3 +.Lend_\@: .endm .macro SAVE_AND_SWITCH_TO_KERNEL_CR3 scratch_reg:req save_reg:req + STATIC_JUMP_IF_FALSE .Ldone_\@, kaiser_enabled_key, def=1 movq %cr3, %r\scratch_reg movq %r\scratch_reg, \save_reg /* @@ -244,11 +249,13 @@ For 32-bit we have the following convent .endm .macro RESTORE_CR3 save_reg:req + STATIC_JUMP_IF_FALSE .Lend_\@, kaiser_enabled_key, def=1 /* * The CR3 write could be avoided when not changing its value, * but would require a CR3 read *and* a scratch register. */ movq \save_reg, %cr3 +.Lend_\@: .endm #else /* CONFIG_KAISER=n: */ --- a/arch/x86/include/asm/kaiser.h +++ b/arch/x86/include/asm/kaiser.h @@ -56,6 +56,16 @@ extern void kaiser_remove_mapping(unsign */ extern void kaiser_init(void); +/* True if kaiser is enabled at boot time */ +extern struct static_key_true kaiser_enabled_key; +extern bool kaiser_enabled; +extern void kaiser_check_cmdline(void); + +#else /* CONFIG_KAISER */ + +#define kaiser_enabled (false) +static inline void kaiser_check_cmdline(void) { } + #endif #endif /* __ASSEMBLY__ */ --- a/arch/x86/include/asm/pgtable_64.h +++ b/arch/x86/include/asm/pgtable_64.h @@ -175,6 +175,9 @@ static inline p4d_t *shadow_to_kernel_p4 { return ptr_clear_bit(p4dp, KAISER_PGTABLE_SWITCH_BIT); } + +extern bool kaiser_enabled; + #endif /* CONFIG_KAISER */ /* @@ -208,6 +211,9 @@ static inline bool pgd_userspace_access( static inline pgd_t kaiser_set_shadow_pgd(pgd_t *pgdp, pgd_t pgd) { #ifdef CONFIG_KAISER + if (!kaiser_enabled) + return pgd; + if (pgd_userspace_access(pgd)) { if (pgdp_maps_userspace(pgdp)) { /* --- a/arch/x86/mm/dump_pagetables.c +++ b/arch/x86/mm/dump_pagetables.c @@ -20,6 +20,7 @@ #include #include +#include /* * The dumper groups pagetable entries of the same type into one, and for @@ -503,7 +504,7 @@ void ptdump_walk_pgd_level(struct seq_fi void ptdump_walk_pgd_level_debugfs(struct seq_file *m, pgd_t *pgd, bool shadow) { - if (shadow) + if (shadow && kaiser_enabled) pgd += PTRS_PER_PGD; ptdump_walk_pgd_level_core(m, pgd, false, false); } @@ -514,6 +515,8 @@ void ptdump_walk_shadow_pgd_level_checkw #ifdef CONFIG_KAISER pgd_t *pgd = (pgd_t *) &init_top_pgt; + if (!kaiser_enabled) + return; pr_info("x86/mm: Checking shadow page tables\n"); pgd += PTRS_PER_PGD; ptdump_walk_pgd_level_core(NULL, pgd, true, false); --- a/arch/x86/mm/init.c +++ b/arch/x86/mm/init.c @@ -20,6 +20,7 @@ #include #include #include +#include /* * We need to define the tracepoints somewhere, and tlb.c @@ -163,9 +164,8 @@ static int page_size_mask; static void enable_global_pages(void) { -#ifndef CONFIG_KAISER - __supported_pte_mask |= _PAGE_GLOBAL; -#endif + if (!kaiser_enabled) + __supported_pte_mask |= _PAGE_GLOBAL; } static void __init probe_page_size_mask(void) @@ -656,6 +656,7 @@ void __init init_mem_mapping(void) { unsigned long end; + kaiser_check_cmdline(); probe_page_size_mask(); setup_pcid(); --- a/arch/x86/mm/kaiser.c +++ b/arch/x86/mm/kaiser.c @@ -34,6 +34,7 @@ #include #include +#include #include #include #include @@ -44,6 +45,16 @@ static pteval_t kaiser_pte_mask __ro_after_init = ~(_PAGE_NX | _PAGE_GLOBAL); +/* Global flag for boot time kaiser enable/disable */ +bool kaiser_enabled __ro_after_init = true; +DEFINE_STATIC_KEY_TRUE(kaiser_enabled_key); + +void __init kaiser_check_cmdline(void) +{ + if (cmdline_find_option_bool(boot_command_line, "nokaiser")) + kaiser_enabled = false; +} + /* * At runtime, the only things we map are some things for CPU * hotplug, and stacks for new processes. No two CPUs will ever @@ -252,6 +263,9 @@ int kaiser_add_user_map(const void *__st unsigned long target_address; pte_t *pte; + if (!kaiser_enabled) + return 0; + /* Clear not supported bits */ flags &= kaiser_pte_mask; @@ -402,6 +416,9 @@ void __init kaiser_init(void) { int cpu; + if (!kaiser_enabled) + return; + kaiser_init_all_pgds(); for_each_possible_cpu(cpu) { @@ -436,6 +453,16 @@ void __init kaiser_init(void) kaiser_add_mapping_cpu_entry(0); } +static int __init kaiser_boottime_control(void) +{ + if (!kaiser_enabled) { + static_branch_disable(&kaiser_enabled_key); + pr_info("kaiser: Disabled on command line\n"); + } + return 0; +} +subsys_initcall(kaiser_boottime_control); + int kaiser_add_mapping(unsigned long addr, unsigned long size, unsigned long flags) { @@ -446,6 +473,9 @@ void kaiser_remove_mapping(unsigned long { unsigned long addr; + if (!kaiser_enabled) + return; + /* The shadow page tables always use small pages: */ for (addr = start; addr < start + size; addr += PAGE_SIZE) { /* --- a/security/Kconfig +++ b/security/Kconfig @@ -56,7 +56,7 @@ config SECURITY_NETWORK config KAISER bool "Remove the kernel mapping in user mode" - depends on X86_64 && SMP && !PARAVIRT + depends on X86_64 && SMP && !PARAVIRT && JUMP_LABEL help This feature reduces the number of hardware side channels by ensuring that the majority of kernel addresses are not mapped