From ab14de6c37fae22911ba99f4171613e6d758050b Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 5 Feb 2007 21:18:37 +0100 Subject: [S390] Convert memory detection into C code. Hopefully this will make it more maintainable and less error prone. Code makes use of search_exception_tables(). Since it calls this function before the kernel exeception table is sorted, there is an early call to sort_main_extable(). This way it's easy to use the already present infrastructure of fixup sections. Also this would allows to easily convert the rest of head[31|64].S into C code. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky --- arch/s390/kernel/early.c | 307 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 307 insertions(+) create mode 100644 arch/s390/kernel/early.c (limited to 'arch/s390/kernel/early.c') diff --git a/arch/s390/kernel/early.c b/arch/s390/kernel/early.c new file mode 100644 index 00000000000..40dd47970a3 --- /dev/null +++ b/arch/s390/kernel/early.c @@ -0,0 +1,307 @@ +/* + * arch/s390/kernel/early.c + * + * Copyright IBM Corp. 2007 + * Author(s): Hongjie Yang , + * Heiko Carstens + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +/* + * Create a Kernel NSS if the SAVESYS= parameter is defined + */ +#define DEFSYS_CMD_SIZE 96 +#define SAVESYS_CMD_SIZE 32 + +extern int _eshared; +char kernel_nss_name[NSS_NAME_SIZE + 1]; + +#ifdef CONFIG_SHARED_KERNEL +static noinline __init void create_kernel_nss(void) +{ + unsigned int i, stext_pfn, eshared_pfn, end_pfn, min_size; +#ifdef CONFIG_BLK_DEV_INITRD + unsigned int sinitrd_pfn, einitrd_pfn; +#endif + int response; + char *savesys_ptr; + char upper_command_line[COMMAND_LINE_SIZE]; + char defsys_cmd[DEFSYS_CMD_SIZE]; + char savesys_cmd[SAVESYS_CMD_SIZE]; + + /* Do nothing if we are not running under VM */ + if (!MACHINE_IS_VM) + return; + + /* Convert COMMAND_LINE to upper case */ + for (i = 0; i < strlen(COMMAND_LINE); i++) + upper_command_line[i] = toupper(COMMAND_LINE[i]); + + savesys_ptr = strstr(upper_command_line, "SAVESYS="); + + if (!savesys_ptr) + return; + + savesys_ptr += 8; /* Point to the beginning of the NSS name */ + for (i = 0; i < NSS_NAME_SIZE; i++) { + if (savesys_ptr[i] == ' ' || savesys_ptr[i] == '\0') + break; + kernel_nss_name[i] = savesys_ptr[i]; + } + + stext_pfn = PFN_DOWN(__pa(&_stext)); + eshared_pfn = PFN_DOWN(__pa(&_eshared)); + end_pfn = PFN_UP(__pa(&_end)); + min_size = end_pfn << 2; + + sprintf(defsys_cmd, "DEFSYS %s 00000-%.5X EW %.5X-%.5X SR %.5X-%.5X", + kernel_nss_name, stext_pfn - 1, stext_pfn, eshared_pfn - 1, + eshared_pfn, end_pfn); + +#ifdef CONFIG_BLK_DEV_INITRD + if (INITRD_START && INITRD_SIZE) { + sinitrd_pfn = PFN_DOWN(__pa(INITRD_START)); + einitrd_pfn = PFN_UP(__pa(INITRD_START + INITRD_SIZE)); + min_size = einitrd_pfn << 2; + sprintf(defsys_cmd, "%s EW %.5X-%.5X", defsys_cmd, + sinitrd_pfn, einitrd_pfn); + } +#endif + + sprintf(defsys_cmd, "%s EW MINSIZE=%.7iK", defsys_cmd, min_size); + sprintf(savesys_cmd, "SAVESYS %s \n IPL %s", + kernel_nss_name, kernel_nss_name); + + __cpcmd(defsys_cmd, NULL, 0, &response); + + if (response != 0) + return; + + __cpcmd(savesys_cmd, NULL, 0, &response); + + if (response != strlen(savesys_cmd)) + return; + + ipl_flags = IPL_NSS_VALID; +} + +#else /* CONFIG_SHARED_KERNEL */ + +static inline void create_kernel_nss(void) { } + +#endif /* CONFIG_SHARED_KERNEL */ + +/* + * Clear bss memory + */ +static noinline __init void clear_bss_section(void) +{ + memset(__bss_start, 0, _end - __bss_start); +} + +/* + * Initialize storage key for kernel pages + */ +static noinline __init void init_kernel_storage_key(void) +{ + unsigned long end_pfn, init_pfn; + + end_pfn = PFN_UP(__pa(&_end)); + + for (init_pfn = 0 ; init_pfn < end_pfn; init_pfn++) + page_set_storage_key(init_pfn << PAGE_SHIFT, PAGE_DEFAULT_KEY); +} + +static noinline __init void detect_machine_type(void) +{ + struct cpuinfo_S390 *cpuinfo = &S390_lowcore.cpu_data; + + asm volatile("stidp %0" : "=m" (S390_lowcore.cpu_data.cpu_id)); + + /* Running under z/VM ? */ + if (cpuinfo->cpu_id.version == 0xff) + machine_flags |= 1; + + /* Running on a P/390 ? */ + if (cpuinfo->cpu_id.machine == 0x7490) + machine_flags |= 4; +} + +static noinline __init int memory_fast_detect(void) +{ + + unsigned long val0 = 0; + unsigned long val1 = 0xc; + int ret = -ENOSYS; + + if (ipl_flags & IPL_NSS_VALID) + return -ENOSYS; + + asm volatile( + " diag %1,%2,0x260\n" + "0: lhi %0,0\n" + "1:\n" + EX_TABLE(0b,1b) + : "+d" (ret), "+d" (val0), "+d" (val1) : : "cc"); + + if (ret || val0 != val1) + return -ENOSYS; + + memory_chunk[0].size = val0; + return 0; +} + +#define ADDR2G (1UL << 31) + +static noinline __init unsigned long sclp_memory_detect(void) +{ + struct sclp_readinfo_sccb *sccb; + unsigned long long memsize; + + sccb = &s390_readinfo_sccb; + + if (sccb->header.response_code != 0x10) + return 0; + + if (sccb->rnsize) + memsize = sccb->rnsize << 20; + else + memsize = sccb->rnsize2 << 20; + if (sccb->rnmax) + memsize *= sccb->rnmax; + else + memsize *= sccb->rnmax2; +#ifndef CONFIG_64BIT + /* + * Can't deal with more than 2G in 31 bit addressing mode, so + * limit the value in order to avoid strange side effects. + */ + if (memsize > ADDR2G) + memsize = ADDR2G; +#endif + return (unsigned long) memsize; +} + +static inline __init unsigned long __tprot(unsigned long addr) +{ + int cc = -1; + + asm volatile( + " tprot 0(%1),0\n" + "0: ipm %0\n" + " srl %0,28\n" + "1:\n" + EX_TABLE(0b,1b) + : "+d" (cc) : "a" (addr) : "cc"); + return (unsigned long)cc; +} + +/* Checking memory in 128KB increments. */ +#define CHUNK_INCR (1UL << 17) + +static noinline __init void find_memory_chunks(unsigned long memsize) +{ + unsigned long addr = 0, old_addr = 0; + unsigned long old_cc = CHUNK_READ_WRITE; + unsigned long cc; + int chunk = 0; + + while (chunk < MEMORY_CHUNKS) { + cc = __tprot(addr); + while (cc == old_cc) { + addr += CHUNK_INCR; + cc = __tprot(addr); +#ifndef CONFIG_64BIT + if (addr == ADDR2G) + break; +#endif + } + + if (old_addr != addr && + (old_cc == CHUNK_READ_WRITE || old_cc == CHUNK_READ_ONLY)) { + memory_chunk[chunk].addr = old_addr; + memory_chunk[chunk].size = addr - old_addr; + memory_chunk[chunk].type = old_cc; + chunk++; + } + + old_addr = addr; + old_cc = cc; + +#ifndef CONFIG_64BIT + if (addr == ADDR2G) + break; +#endif + /* + * Finish memory detection at the first hole, unless + * - we reached the hsa -> skip it. + * - we know there must be more. + */ + if (cc == -1UL && !memsize && old_addr != ADDR2G) + break; + if (memsize && addr >= memsize) + break; + } +} + +static __init void early_pgm_check_handler(void) +{ + unsigned long addr; + const struct exception_table_entry *fixup; + + addr = S390_lowcore.program_old_psw.addr; + fixup = search_exception_tables(addr & PSW_ADDR_INSN); + if (!fixup) + disabled_wait(0); + S390_lowcore.program_old_psw.addr = fixup->fixup | PSW_ADDR_AMODE; +} + +static noinline __init void setup_lowcore_early(void) +{ + psw_t psw; + + psw.mask = PSW_BASE_BITS | PSW_DEFAULT_KEY; + psw.addr = PSW_ADDR_AMODE | (unsigned long) s390_base_ext_handler; + S390_lowcore.external_new_psw = psw; + psw.addr = PSW_ADDR_AMODE | (unsigned long) s390_base_pgm_handler; + S390_lowcore.program_new_psw = psw; + s390_base_pgm_handler_fn = early_pgm_check_handler; +} + +/* + * Save ipl parameters, clear bss memory, initialize storage keys + * and create a kernel NSS at startup if the SAVESYS= parm is defined + */ +void __init startup_init(void) +{ + unsigned long memsize; + + ipl_save_parameters(); + clear_bss_section(); + init_kernel_storage_key(); + lockdep_init(); + lockdep_off(); + detect_machine_type(); + create_kernel_nss(); + sort_main_extable(); + setup_lowcore_early(); + sclp_readinfo_early(); + memsize = sclp_memory_detect(); + if (memory_fast_detect() < 0) + find_memory_chunks(memsize); + lockdep_on(); +} -- cgit v1.2.3-70-g09d2 From 162e006ef59266b9ebf34e3d15ca1f3d9ee956d7 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Mon, 5 Feb 2007 21:18:41 +0100 Subject: [S390] Mark kernel text section read-only. Set read-only flag in the page table entries for the kernel image text section. This will catch all instruction caused corruptions withing the text section. Instruction replacement via kprobes still works, since it bypasses now dynamic address translation. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky --- arch/s390/kernel/early.c | 1 - arch/s390/kernel/kprobes.c | 27 +++++++++++++++++++++++---- arch/s390/kernel/vmlinux.lds.S | 3 ++- arch/s390/mm/init.c | 9 ++++----- include/asm-s390/sections.h | 2 ++ 5 files changed, 31 insertions(+), 11 deletions(-) (limited to 'arch/s390/kernel/early.c') diff --git a/arch/s390/kernel/early.c b/arch/s390/kernel/early.c index 40dd47970a3..e518dd53eff 100644 --- a/arch/s390/kernel/early.c +++ b/arch/s390/kernel/early.c @@ -27,7 +27,6 @@ #define DEFSYS_CMD_SIZE 96 #define SAVESYS_CMD_SIZE 32 -extern int _eshared; char kernel_nss_name[NSS_NAME_SIZE + 1]; #ifdef CONFIG_SHARED_KERNEL diff --git a/arch/s390/kernel/kprobes.c b/arch/s390/kernel/kprobes.c index b2e1dc89a8c..a466bab6677 100644 --- a/arch/s390/kernel/kprobes.c +++ b/arch/s390/kernel/kprobes.c @@ -155,15 +155,34 @@ void __kprobes get_instruction_type(struct arch_specific_insn *ainsn) static int __kprobes swap_instruction(void *aref) { struct ins_replace_args *args = aref; + u32 *addr; + u32 instr; int err = -EFAULT; + /* + * Text segment is read-only, hence we use stura to bypass dynamic + * address translation to exchange the instruction. Since stura + * always operates on four bytes, but we only want to exchange two + * bytes do some calculations to get things right. In addition we + * shall not cross any page boundaries (vmalloc area!) when writing + * the new instruction. + */ + addr = (u32 *)ALIGN((unsigned long)args->ptr, 4); + if ((unsigned long)args->ptr & 2) + instr = ((*addr) & 0xffff0000) | args->new; + else + instr = ((*addr) & 0x0000ffff) | args->new << 16; + asm volatile( - "0: mvc 0(2,%2),0(%3)\n" - "1: la %0,0\n" + " lra %1,0(%1)\n" + "0: stura %2,%1\n" + "1: la %0,0\n" "2:\n" EX_TABLE(0b,2b) - : "+d" (err), "=m" (*args->ptr) - : "a" (args->ptr), "a" (&args->new), "m" (args->new)); + : "+d" (err) + : "a" (addr), "d" (instr) + : "memory", "cc"); + return err; } diff --git a/arch/s390/kernel/vmlinux.lds.S b/arch/s390/kernel/vmlinux.lds.S index 8fedb1f9fc9..a4890739252 100644 --- a/arch/s390/kernel/vmlinux.lds.S +++ b/arch/s390/kernel/vmlinux.lds.S @@ -35,9 +35,10 @@ SECTIONS #ifdef CONFIG_SHARED_KERNEL . = ALIGN(1048576); /* VM shared segments are 1MB aligned */ +#endif + . = ALIGN(4096); _eshared = .; /* End of shareable data */ -#endif . = ALIGN(16); /* Exception table */ __start___ex_table = .; diff --git a/arch/s390/mm/init.c b/arch/s390/mm/init.c index 162a338a557..b3e7c45efb6 100644 --- a/arch/s390/mm/init.c +++ b/arch/s390/mm/init.c @@ -26,7 +26,6 @@ #include #include #include - #include #include #include @@ -96,8 +95,8 @@ static void __init setup_ro_region(void) pte_t new_pte; unsigned long address, end; - address = ((unsigned long)&__start_rodata) & PAGE_MASK; - end = PFN_ALIGN((unsigned long)&__end_rodata); + address = ((unsigned long)&_stext) & PAGE_MASK; + end = PFN_ALIGN((unsigned long)&_eshared); for (; address < end; address += PAGE_SIZE) { pgd = pgd_offset_k(address); @@ -173,8 +172,8 @@ void __init mem_init(void) datasize >>10, initsize >> 10); printk("Write protected kernel read-only data: %#lx - %#lx\n", - (unsigned long)&__start_rodata, - PFN_ALIGN((unsigned long)&__end_rodata) - 1); + (unsigned long)&_stext, + PFN_ALIGN((unsigned long)&_eshared) - 1); } void free_initmem(void) diff --git a/include/asm-s390/sections.h b/include/asm-s390/sections.h index 3a0b8ffeab7..1c5a2c4ccda 100644 --- a/include/asm-s390/sections.h +++ b/include/asm-s390/sections.h @@ -3,4 +3,6 @@ #include +extern char _eshared[]; + #endif -- cgit v1.2.3-70-g09d2