Exclude areas not assigned to Mckernel from direct map of all phys. memory

It's enabled by adding -s to mcreboot.sh.

Cherry-pick of the following commit:

commit b5c13ce51a5a4926c2cf11c817cd0d369ac4402d
Author: Katsuya Horigome <katsuya.horigome.rj@ps.hitachi-solutions.com>
Date:   Mon Nov 20 09:40:41 2017 +0900

    Include measures to prevent memory destruction on Linux side (This is rebase commit for merging to development+hfi)

arch/x86_64/kernel/memory.c

@@ -39,6 +39,8 @@ extern char _head[], _end[];
 
 extern unsigned long x86_kernel_phys_base;
 
+int safe_kernel_map = 0;
+
 /* Arch specific early allocation routine */
 void *early_alloc_pages(int nr_pages)
 {
@@ -2545,22 +2547,54 @@ static void init_linux_kernel_mapping(struct page_table *pt)
 {
 	unsigned long map_start, map_end, phys;
 	void *virt;
+	int nr_memory_chunks, chunk_id, numa_id;
 
-	/* Map 2 TB for now */
+	/* In case of safe_kernel_map option (safe_kernel_map == 1),
-	map_start = 0;
+	 * processing to prevent destruction of the memory area on Linux side
-	map_end = 0x20000000000;
+	 * is executed */
+	if (safe_kernel_map == 0) {
+		kprintf("Straight-map entire physical memory\n");
 
-	virt = (void *)LINUX_PAGE_OFFSET;
+		/* Map 2 TB for now */
+		map_start = 0;
+		map_end = 0x20000000000;
 
-	kprintf("Linux kernel virtual: 0x%lx - 0x%lx -> 0x%lx - 0x%lx\n",
+		virt = (void *)LINUX_PAGE_OFFSET;
-		LINUX_PAGE_OFFSET, LINUX_PAGE_OFFSET + map_end, 0, map_end);
 
-	for (phys = map_start; phys < map_end; phs += LARGE_PAGE_SIZE) {
+		kprintf("Linux kernel virtual: 0x%lx - 0x%lx -> 0x%lx - 0x%lx\n",
-		if (set_pt_large_page(pt, virt, phys, PTATTR_WRITABLE) != 0) {
+			LINUX_PAGE_OFFSET, LINUX_PAGE_OFFSET + map_end, 0, map_end);
-			kprintf("%s: error setting mapping for 0x%lx\n",
+
-				__func__, virt);
+		for (phys = map_start; phys < map_end; phys += LARGE_PAGE_SIZE) {
+			if (set_pt_large_page(pt, virt, phys, PTATTR_WRITABLE) != 0) {
+				kprintf("%s: error setting mapping for 0x%lx\n", __FUNCTION__, virt);
+			}
+			virt += LARGE_PAGE_SIZE;
+		}
+	} else {
+		kprintf("Straight-map physical memory areas allocated to McKernel\n");
+
+		nr_memory_chunks = ihk_mc_get_nr_memory_chunks();
+		if (nr_memory_chunks == 0) {
+			kprintf("%s: ERROR: No memory chunk available.\n", __FUNCTION__);
+			return;
+		}
+
+		for (chunk_id = 0; chunk_id < nr_memory_chunks; chunk_id++) {
+			if (ihk_mc_get_memory_chunk(chunk_id, &map_start, &map_end, &numa_id)) {
+				kprintf("%s: ERROR: Memory chunk id (%d) out of range.\n", __FUNCTION__, chunk_id);
+				continue;
+			}
+
+			dkprintf("Linux kernel virtual: 0x%lx - 0x%lx -> 0x%lx - 0x%lx\n",
+					 LINUX_PAGE_OFFSET + map_start, LINUX_PAGE_OFFSET + map_end, map_start, map_end);
+
+			virt = (void *)(LINUX_PAGE_OFFSET + map_start);
+			for (phys = map_start; phys < map_end; phys += LARGE_PAGE_SIZE, virt += LARGE_PAGE_SIZE) {
+				if (set_pt_large_page(pt, virt, phys, PTATTR_WRITABLE) != 0) {
+					kprintf("%s: set_pt_large_page() failed for 0x%lx\n", __FUNCTION__, virt);
+				}
+			}
 		}
-		virt += LARGE_PAGE_SIZE;
 	}
 }
 

arch/x86_64/tools/mcreboot-smp-x86.sh.in

@@ -45,11 +45,12 @@ fi
 
 turbo=""
 ihk_irq=""
+safe_kernel_map=""
 umask_old=`umask`
 idle_halt=""
 allow_oversubscribe=""
 
-while getopts :tk:c:m:o:f:r:q:i:d:e:hO OPT
+while getopts :stk:c:m:o:f:r:q:i:d:e:hO OPT
 do
 	case ${OPT} in
 	f)	facility=${OPTARG}
@@ -62,6 +63,8 @@ do
 		;;
 	m) mem=${OPTARG}
 		;;
+	s) safe_kernel_map="safe_kernel_map"
+		;;
 	r) ikc_map=${OPTARG}
 		;;
 	q) ihk_irq=${OPTARG}
@@ -446,7 +449,7 @@ if ! ${SBINDIR}/ihkosctl 0 load ${KERNDIR}/mckernel.img; then
 fi
 
 # Set kernel arguments
-if ! ${SBINDIR}/ihkosctl 0 kargs "hidos $turbo $idle_halt dump_level=${DUMP_LEVEL} $extra_kopts $allow_oversubscribe"; then
+if ! ${SBINDIR}/ihkosctl 0 kargs "hidos $turbo $safe_kernel_map $idle_halt dump_level=${DUMP_LEVEL} $extra_kopts $allow_oversubscribe"; then
 	echo "error: setting kernel arguments" >&2
 	error_exit "os_created"
 fi

kernel/init.c

@@ -123,6 +123,8 @@ char *find_command_line(char *name)
 	return strstr(cmdline, name);
 }
 
+extern int safe_kernel_map;
+
 static void parse_kargs(void)
 {
 	char *ptr;
@@ -143,6 +145,11 @@ static void parse_kargs(void)
 	}
 	ihk_mc_set_dump_level(dump_level);
 
+	ptr = find_command_line("safe_kernel_map");
+	if (ptr) {
+		safe_kernel_map = 1;
+	}
+
 	/* idle_halt option */
 	ptr = find_command_line("idle_halt");
 	if (ptr) {

test/mem_dest_prev/LTP_testcase.txt

@@ -0,0 +1,100 @@
+brk01
+clone01
+clone03
+clone04
+clone06
+clone07
+close01
+close02
+dup01
+dup02
+dup03
+dup04
+dup05
+dup06
+dup07
+fork01
+fork02
+fork03
+fork04
+fork07
+fork08
+fork09
+fork10
+fork11
+mbind01
+mem01
+mem02
+mem03
+memcpy01
+memfd_create02
+memset01
+mkdir01
+mkdir08
+mkdirat01
+mknodat01
+mmap001
+mmap01
+mmap02
+mmap03
+mmap04
+mmap06
+mmap07
+mmap08
+mmap09
+mmap12
+mmapstress02
+mmapstress04
+mmapstress05
+mremap01
+mremap05
+open01
+open03
+open04
+open06
+open07
+open09
+open13
+poll01
+posix_fadvise01
+read01
+read02
+read03
+read04
+sbrk01
+sbrk02
+sendfile02
+sendfile03
+sendfile04
+sendfile05
+sendfile06
+sendfile07
+sendfile08
+sendfile09
+semctl01
+semctl03
+semctl05
+socket01
+socket02
+stream01
+stream02
+stream03
+stream04
+stream05
+unlink05
+unlink06
+unlink07
+unlink08
+vfork01
+vfork02
+vma01
+vmsplice01
+vmsplice02
+write01
+write03
+write04
+write05
+writetest
+writev01
+writev02
+writev07

test/mem_dest_prev/README

@@ -0,0 +1,25 @@
+===================
+Advance preparation
+===================
+1)Implement patch of test_memtest_destroy.patch
+cd mckernel
+patch -p0 <  test_memtest_destroy.patch
+make
+make install
+
+
+2)Compile command execution processing
+cd mckernel/test/mem_dest_prev/mcexec_test_proc/
+make
+
+3)Write the LTP path to LTP_DIR in the configuration file
+vi config
+
+  ex) LTP_DIR=$HOME/test/mem_dest_prev/ltp/testcases/bin/
+
+
+==========
+How to run
+==========
+./go_test_McKernal.sh
+

test/mem_dest_prev/config

@@ -0,0 +1,13 @@
+MCMOD_DIR=$HOME/ppos
+LTP_DIR=$HOME/test/mem_dest_prev/ltp/testcases/bin/
+LTP_TESTCASE_FILE=LTP_testcase.txt
+MCRBT_OPT_LTP="-m 3G@0,3G@1 -s"
+USR_PROC="mcexec_test_proc/memtest_destroy"
+OS_IDX=0
+
+export MCMOD_DIR
+export LTP_DIR
+export LTP_TESTCASE_FILE
+export MCRBT_OPT_LTP
+export USR_PROC
+export OS_IDX

test/mem_dest_prev/go_test_McKernal.sh

@@ -0,0 +1,101 @@
+#!/bin/sh
+
+# read config
+source ./config
+
+#logfile="./result/test_result.log"
+
+# mcexec processのkill
+./utils/kill_mcexec.sh &> /dev/null
+
+for test_case in `ls -1 ./testcases/*.txt`
+do
+	# read testcase param
+	source ${test_case}
+	case_name=`basename ${test_case} .txt`
+	echo "####################"
+	echo "Test No:${case_name}"
+
+	# Out-of-range address Test(Before correspondence)
+	echo ">>> Out-of-range address Test(Before correspondence) Start"
+
+	# stop mckernel
+	sudo ${MCMOD_DIR}/sbin/mcstop+release.sh
+	sleep 1
+	# boot mckernel
+	echo "${MCMOD_DIR}/sbin/mcreboot.sh ${MCRBT_OPT_BEFORE%,}"
+	sudo ${MCMOD_DIR}/sbin/mcreboot.sh ${MCRBT_OPT_BEFORE%,}
+	sleep 1
+
+	echo "    ${MCMOD_DIR}/bin/mcexec ${USR_PROC}"
+	timeout -sKILL 5 ${MCMOD_DIR}/bin/mcexec ${USR_PROC}
+	STATUS=$?
+
+	echo "${MCMOD_DIR}/sbin/ihkosctl ${OS_IDX} kmsg"
+	sudo ${MCMOD_DIR}/sbin/ihkosctl ${OS_IDX} kmsg
+
+	if [ "$STATUS" -ne 21 ];
+	then
+		echo ">>> Out-of-range address Test End(Timeout!!!)"
+	else
+		echo ">>> Out-of-range address Test End"
+	fi
+
+	# Out-of-range address Test(After correspondence)
+	echo ">>> Out-of-range address(After correspondence) Test Start"
+
+	# stop mckernel
+	sudo ${MCMOD_DIR}/sbin/mcstop+release.sh
+	sleep 1
+	# boot mckernel
+	echo "${MCMOD_DIR}/sbin/mcreboot.sh ${MCRBT_OPT_AFTER%,}"
+	sudo ${MCMOD_DIR}/sbin/mcreboot.sh ${MCRBT_OPT_AFTER%,}
+	sleep 1
+
+	echo "    ${MCMOD_DIR}/bin/mcexec ${USR_PROC}"
+	timeout -sKILL 5 ${MCMOD_DIR}/bin/mcexec ${USR_PROC}
+	STATUS=$?
+
+	echo "${MCMOD_DIR}/sbin/ihkosctl ${OS_IDX} kmsg"
+	sudo ${MCMOD_DIR}/sbin/ihkosctl ${OS_IDX} kmsg
+
+	if [ "$STATUS" -ne 21 ];
+	then
+		echo ">>> Out-of-range address Test End(Timeout!!!)"
+	else
+		echo ">>> Out-of-range address Test End"
+	fi
+done
+
+### LTP START ##################################################
+# stop mckernel
+sudo ${MCMOD_DIR}/sbin/mcstop+release.sh
+sleep 1
+
+# boot mckernel
+echo "${MCMOD_DIR}/sbin/mcreboot.sh ${MCRBT_OPT_LTP%,}"
+sudo ${MCMOD_DIR}/sbin/mcreboot.sh ${MCRBT_OPT_LTP%,}
+sleep 1
+
+if [ ! -e "/dev/mcos0" ]; then
+	echo "Error: failed to mcreboot"
+	exit 1
+fi
+
+
+TEST_NUM=`wc -l ${LTP_TESTCASE_FILE} | awk '{print $1}'`
+echo ">>> LTP Test Start( $TEST_NUM counts )"
+
+# exec mckernel test program
+COUNT=0
+while read line
+do 
+	((COUNT++))
+	echo "$COUNT:${MCMOD_DIR}/bin/mcexec ${LTP_DIR}$line"
+#	${MCMOD_DIR}/bin/mcexec ${LTP_DIR}$line &>> ${logfile}
+	${MCMOD_DIR}/bin/mcexec ${LTP_DIR}$line
+done < ${LTP_TESTCASE_FILE}
+
+echo ">>> LTP Test End"
+### LTP END ####################################################
+

test/mem_dest_prev/mcexec_test_proc/Makefile

@@ -0,0 +1,7 @@
+OBJS = memtest_destroy
+
+all:$(OBJS)
+
+clean:
+	rm $(OBJS)
+

test/mem_dest_prev/mcexec_test_proc/memtest_destroy.c

@@ -0,0 +1,13 @@
+#include <stdio.h>
+#define _GNU_SOURCE
+#include <unistd.h>
+#include <sys/syscall.h>
+
+main() {
+	int rst = 0;
+	
+	rst = syscall(950);
+	printf("mem_destroy result:%d\n",rst);
+
+	return;
+}

test/mem_dest_prev/test_memtest_destroy.patch

@@ -0,0 +1,50 @@
+diff --git arch/x86/kernel/include/syscall_list.h arch/x86/kernel/include/syscall_list.h
+index 42d1e2e..f5769b8 100644
+--- arch/x86/kernel/include/syscall_list.h
++++ arch/x86/kernel/include/syscall_list.h
+@@ -156,5 +156,7 @@ SYSCALL_HANDLED(__NR_profile, profile)
+ SYSCALL_HANDLED(730, util_migrate_inter_kernel)
+ SYSCALL_HANDLED(731, util_indicate_clone)
+ SYSCALL_HANDLED(732, get_system)
++/* McKernel Specific */
++SYSCALL_HANDLED(950, mem_destroy)
+ 
+ /**** End of File ****/
+diff --git arch/x86/kernel/syscall.c arch/x86/kernel/syscall.c
+index 2260b66..e96776a 100644
+--- arch/x86/kernel/syscall.c
++++ arch/x86/kernel/syscall.c
+@@ -1887,4 +1887,33 @@ save_uctx(void *uctx, struct x86_user_context *regs)
+ 	ctx->fregsize = 0;
+ }
+ 
++
++#define ADD_ADDR_VAL 0x400
++SYSCALL_DECLARE(mem_destroy)
++{
++	int rst = 0;
++	int mem_chunks_num, chunk_id, get_numa_id;
++	unsigned long get_start, get_end;
++	unsigned long *addr;
++
++	mem_chunks_num = ihk_mc_get_nr_memory_chunks();
++	kprintf("%s: memory chunk %d.\n", __FUNCTION__, mem_chunks_num);
++
++	for (chunk_id = 0; chunk_id < mem_chunks_num; chunk_id++) {
++		rst = ihk_mc_get_memory_chunk(chunk_id, &get_start, &get_end, &get_numa_id);
++		kprintf("%s: mem chunk[%d] numa ID(%d)\n"
++			,__FUNCTION__ ,chunk_id ,get_numa_id);
++		kprintf("    phys(0x%lx - 0x%lx) virt(0x%lx - 0x%lx)\n"
++			,get_start ,get_end ,phys_to_virt(get_start) ,phys_to_virt(get_end));
++	}
++
++	addr  = phys_to_virt(get_end + ADD_ADDR_VAL);
++#if 1 
++	*addr = 0x1;
++#endif     
++	kprintf("%s: Address out of range  0x%lx(val:%d)\n",__FUNCTION__ ,addr ,*addr);
++
++	return rst;
++}
++
+ /*** End of File ***/

test/mem_dest_prev/testcases/0001.txt

@@ -0,0 +1,2 @@
+MCRBT_OPT_BEFORE="-m `./utils/gen_mem_chunks.sh "0 1" 3G 1`"
+MCRBT_OPT_AFTER="-m `./utils/gen_mem_chunks.sh "0 1" 3G 1` -s"

test/mem_dest_prev/testcases/0002.txt

@@ -0,0 +1,2 @@
+MCRBT_OPT_BEFORE="-m `./utils/gen_mem_chunks.sh "0 1" 32M 2`"
+MCRBT_OPT_AFTER="-m `./utils/gen_mem_chunks.sh "0 1" 32M 2` -s"

test/mem_dest_prev/testcases/0003.txt

@@ -0,0 +1,2 @@
+MCRBT_OPT_BEFORE="-m `./utils/gen_mem_chunks.sh "0 1" 32M 4`"
+MCRBT_OPT_AFTER="-m `./utils/gen_mem_chunks.sh "0 1" 32M 4` -s"

test/mem_dest_prev/testcases/0004.txt

@@ -0,0 +1,2 @@
+MCRBT_OPT_BEFORE="-m `./utils/gen_mem_chunks.sh "0 1" 32M 8`"
+MCRBT_OPT_AFTER="-m `./utils/gen_mem_chunks.sh "0 1" 32M 8` -s"

test/mem_dest_prev/testcases/0005.txt

@@ -0,0 +1,2 @@
+MCRBT_OPT_BEFORE="-m `./utils/gen_mem_chunks.sh "0 1" 32M 16`"
+MCRBT_OPT_AFTER="-m `./utils/gen_mem_chunks.sh "0 1" 32M 16` -s"

test/mem_dest_prev/testcases/0006.txt

@@ -0,0 +1,2 @@
+MCRBT_OPT_BEFORE="-m `./utils/gen_mem_chunks.sh "0 1" 32M 32`"
+MCRBT_OPT_AFTER="-m `./utils/gen_mem_chunks.sh "0 1" 32M 32` -s"

test/mem_dest_prev/testcases/0007.txt

@@ -0,0 +1,2 @@
+MCRBT_OPT_BEFORE="-m `./utils/gen_mem_chunks.sh "0 1" 32M 48`"
+MCRBT_OPT_AFTER="-m `./utils/gen_mem_chunks.sh "0 1" 32M 48` -s"

test/mem_dest_prev/testcases/0008.txt

@@ -0,0 +1,2 @@
+MCRBT_OPT_BEFORE="-m `./utils/gen_mem_chunks.sh "0 1" 32M 64`"
+MCRBT_OPT_AFTER="-m `./utils/gen_mem_chunks.sh "0 1" 32M 64` -s"

test/mem_dest_prev/testcases/0009.txt

@@ -0,0 +1,2 @@
+MCRBT_OPT_BEFORE="-m `./utils/gen_mem_chunks.sh "0 1" 32M 96`"
+MCRBT_OPT_AFTER="-m `./utils/gen_mem_chunks.sh "0 1" 32M 96` -s"

test/mem_dest_prev/testcases/0010.txt

@@ -0,0 +1,2 @@
+MCRBT_OPT_BEFORE="-m `./utils/gen_mem_chunks.sh "0 1" 32M 128`"
+MCRBT_OPT_AFTER="-m `./utils/gen_mem_chunks.sh "0 1" 32M 128` -s"

test/mem_dest_prev/utils/gen_mem_chunks.sh

@@ -0,0 +1,16 @@
+#!/bin/sh
+
+NUMAS=$1
+MEM_SIZE=$2
+REP=$3
+CHUNKS=""
+
+for numa in ${NUMAS}
+do
+	for rep in `seq 1 ${REP}`
+	do
+		CHUNKS="${CHUNKS}${MEM_SIZE}@${numa},"
+	done
+done
+
+echo ${CHUNKS%,}

test/mem_dest_prev/utils/kill_mcexec.sh

@@ -0,0 +1,10 @@
+#!/bin/sh
+
+count=`pgrep -c -f 'mcexec '`
+if [ ${count} -gt 0 ]
+then
+	echo "kill process :" ${count}
+	pgrep -l -f 'mcexec '
+	pgrep -f 'mcexec ' | xargs kill -9
+fi
+