From: Siarhei Liakh RO/NX protection for loadable kernel modules Date: Wed, 8 Jul 2009 19:10:34 -0400 This patch is a logical extension of the protection provided by CONFIG_DEBUG_RODATA to LKMs. The protection is provided by splitting module_core and module_init into three logical parts each and setting appropriate page access permissions for each individual section: 1. Code: RO+X 2. RO data: RO+NX 3. RW data: RW+NX In order to achieve proper protection, layout_sections() have been modified to align each of the three parts mentioned above onto page boundary. Next, the corresponding page access permissions are set right before successful exit from load_module(). Further, free_module() and sys_init_module have been modified to set module_core and module_init as RW+NX right before calling module_free(). By default, the original section layout is preserved and RO/NX is enforced only for whole pages of same content. However, when compiled with CONFIG_DEBUG_RODATA=y, the patch will page-align each group of sections to ensure that each page contains only one type of content. v1: Initial proof-of-concept patch. v2: The patch have been re-written to reduce the number of #ifdefs and to make it architecture-agnostic. Code formatting have been corrected also. v3: Opportunistic RO/NX protectiuon is now unconditional. Section page-alignment is enabled when CONFIG_DEBUG_RODATA=y. v4: Removed most macros and improved coding style. v5: Changed page-alignment and RO/NX section size calculation The patch have been developed for Linux 2.6.30 by Siarhei Liakh and Xuxian Jiang . Signed-off-by: Siarhei Liakh Signed-off-by: Xuxian Jiang Acked-by: Arjan van de Ven Signed-off-by: Rusty Russell (fixes) --- include/linux/module.h | 3 + kernel/module.c | 132 ++++++++++++++++++++++++++++++++++++++++++++++++- 2 files changed, 133 insertions(+), 2 deletions(-) diff --git a/include/linux/module.h b/include/linux/module.h --- a/include/linux/module.h +++ b/include/linux/module.h @@ -296,6 +296,9 @@ struct module /* The size of the executable code in each section. */ unsigned int init_text_size, core_text_size; + /* Size of RO sections of the module (text+rodata) */ + unsigned int init_ro_size, core_ro_size; + /* Arch-specific module values */ struct mod_arch_specific arch; diff --git a/kernel/module.c b/kernel/module.c --- a/kernel/module.c +++ b/kernel/module.c @@ -54,6 +54,7 @@ #include #include #include +#include #define CREATE_TRACE_POINTS #include @@ -70,6 +71,22 @@ EXPORT_TRACEPOINT_SYMBOL(module_get); #define ARCH_SHF_SMALL 0 #endif +/* Modules' sections will be aligned on page boundaries + * to ensure complete separation of code and data, but + * only when CONFIG_DEBUG_RODATA=y */ +#ifdef CONFIG_DEBUG_RODATA +#define debug_align(X) ALIGN(X, PAGE_SIZE) +#else +#define debug_align(X) (X) +#endif + +/* Given BASE and SIZE this macro calculates the number of pages the + * memory regions occupies */ +#define NUMBER_OF_PAGES(BASE, SIZE) ((SIZE > 0) ? \ + (PFN_DOWN((unsigned long)BASE + SIZE - 1) - \ + PFN_DOWN((unsigned long)BASE) + 1) \ + : (0UL)) + /* If this is set, the section belongs in the init part of the module */ #define INIT_OFFSET_MASK (1UL << (BITS_PER_LONG-1)) @@ -1501,6 +1518,78 @@ static int __unlink_module(void *_mod) return 0; } +/* LKM RO/NX protection: protect module's text/ro-data + * from modification and any data from execution. + * Siarhei Liakh, Xuxian Jiang */ +static void set_section_ro_nx(void *base, + unsigned long text_size, + unsigned long ro_size, + unsigned long total_size) +{ + /* begin and end PFNs of the current subsection */ + unsigned long begin_pfn; + unsigned long end_pfn; + +#ifdef CONFIG_DEBUG_RODATA + /* Most module_alloc use vmalloc which page-aligns. If not, + * we could be missing protection on first part of module. */ + WARN_ON(offset_in_page((unsigned long)base)); +#endif + + /* Initially, all module sections have RWX permissions*/ + DEBUGP("PROTECTING MODULE SECTION: 0x%lx\n" + " text size: %lu\n" + " ro size: %lu\n" + " total size: %lu\n", + (unsigned long)base, + text_size, ro_size, total_size); + + /* Set RO for module text and RO-data */ + /* Don't protect partial pages. */ + begin_pfn = PFN_UP((unsigned long)base); + end_pfn = PFN_DOWN((unsigned long)base + ro_size); + + /* Set text RO if there are still pages between begin and end */ + if (end_pfn > begin_pfn) { + DEBUGP(" RO: 0x%lx %lu\n", + begin_pfn << PAGE_SHIFT, end_pfn - begin_pfn); + set_memory_ro(begin_pfn << PAGE_SHIFT, end_pfn - begin_pfn); + } + + /* Set NX permissions for module data */ + /* Don't protect partial pages. */ + begin_pfn = PFN_UP((unsigned long)base + text_size); + end_pfn = PFN_DOWN((unsigned long)base + total_size); + + if (end_pfn > begin_pfn) { + DEBUGP(" NX: 0x%lx %lu\n", + begin_pfn << PAGE_SHIFT, end_pfn - begin_pfn); + set_memory_nx(begin_pfn << PAGE_SHIFT, end_pfn - begin_pfn); + } +} + +/* Setting memory back to RW+NX before releasing it */ +void unset_section_ro_nx(struct module *mod, void *module_region) +{ + unsigned long total_pages; + + if (mod->module_core == module_region) { + /* Set core as NX+RW */ + total_pages = NUMBER_OF_PAGES(mod->module_core, mod->core_size); + DEBUGP("Restoring RW+NX for module's CORE: 0x%lx %lu\n", + (unsigned long)mod->module_core, total_pages); + set_memory_nx((unsigned long)mod->module_core, total_pages); + set_memory_rw((unsigned long)mod->module_core, total_pages); + } else if (mod->module_init == module_region) { + /* Set init as NX+RW */ + total_pages = NUMBER_OF_PAGES(mod->module_init, mod->init_size); + DEBUGP("Restoring RW+NX for module's INIT: 0x%lx %lu\n", + (unsigned long)mod->module_init, total_pages); + set_memory_nx((unsigned long)mod->module_init, total_pages); + set_memory_rw((unsigned long)mod->module_init, total_pages); + } +} + /* Free a module, remove from lists, etc (must hold module_mutex). */ static void free_module(struct module *mod) { @@ -1522,6 +1611,7 @@ static void free_module(struct module *m destroy_params(mod->kp, mod->num_kp); /* This may be NULL, but that's OK */ + unset_section_ro_nx(mod, mod->module_init); module_free(mod, mod->module_init); kfree(mod->args); if (mod->percpu) @@ -1534,6 +1624,7 @@ static void free_module(struct module *m lockdep_free_key_range(mod->module_core, mod->core_size); /* Finally, free the core (containing the module structure) */ + unset_section_ro_nx(mod, mod->module_core); module_free(mod, mod->module_core); } @@ -1707,8 +1798,20 @@ static void layout_sections(struct modul s->sh_entsize = get_offset(mod, &mod->core_size, s, i); DEBUGP("\t%s\n", secstrings + s->sh_name); } - if (m == 0) + switch (m) { + case 0: /* executable */ + mod->core_size = debug_align(mod->core_size); mod->core_text_size = mod->core_size; + break; + case 1: /* RO: text and ro-data */ + mod->core_size = debug_align(mod->core_size); + mod->core_ro_size = mod->core_size; + break; + case 3: /* whole module core (executable + RO data + + * RW data + small alloc) */ + mod->core_size = debug_align(mod->core_size); + break; + } } DEBUGP("Init section allocation order:\n"); @@ -1725,8 +1828,20 @@ static void layout_sections(struct modul | INIT_OFFSET_MASK); DEBUGP("\t%s\n", secstrings + s->sh_name); } - if (m == 0) + switch (m) { + case 0: /* executable */ + mod->init_size = debug_align(mod->init_size); mod->init_text_size = mod->init_size; + break; + case 1: /* RO: text and ro-data */ + mod->init_size = debug_align(mod->init_size); + mod->init_ro_size = mod->init_size; + break; + case 3: /* whole module init (executable + RO data + + * RW data + small alloc) */ + mod->init_size = debug_align(mod->init_size); + break; + } } } @@ -2496,6 +2611,18 @@ static noinline struct module *load_modu trace_module_load(mod); + /* Set RO and NX regions for core */ + set_section_ro_nx(mod->module_core, + mod->core_text_size, + mod->core_ro_size, + mod->core_size); + + /* Set RO and NX regions for init */ + set_section_ro_nx(mod->module_init, + mod->init_text_size, + mod->init_ro_size, + mod->init_size); + /* Done! */ return mod; @@ -2611,6 +2738,7 @@ SYSCALL_DEFINE3(init_module, void __user mutex_lock(&module_mutex); /* Drop initial reference. */ module_put(mod); + unset_section_ro_nx(mod, mod->module_init); trim_init_extable(mod); #ifdef CONFIG_KALLSYMS mod->num_symtab = mod->core_num_syms;