move_pages: Fix and support some specs for LTP.

1. When nodes array is NULL, move_pages doesn't move any pages,
 instead will return the node where each page
 currently resides by status array.
2. Check whether all specified node is online or not.

Change-Id: Ie3534997833d797e2a9f595d1107b07d46e1c6cf
Refs: #1523

arch/arm64/kernel/syscall.c

@@ -16,6 +16,7 @@
 #include <uio.h>
 #include <syscall.h>
 #include <rusage_private.h>
+#include <memory.h>
 #include <ihk/debug.h>
 
 void terminate_mcexec(int, int);
@@ -2250,8 +2251,10 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 		case 0:
 			memcpy(mpsr->virt_addr, mpsr->user_virt_addr,
 			       sizeof(void *) * count);
+			if (mpsr->user_nodes) {
 				memcpy(mpsr->nodes, mpsr->user_nodes,
 						sizeof(int) * count);
+			}
 			memset(mpsr->ptep, 0, sizeof(pte_t) * count);
 			memset(mpsr->status, 0, sizeof(int) * count);
 			memset(mpsr->nr_pages, 0, sizeof(int) * count);
@@ -2269,8 +2272,10 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 		case 0:
 			memcpy(mpsr->virt_addr, mpsr->user_virt_addr,
 			       sizeof(void *) * count);
+			if (mpsr->user_nodes) {
 				memcpy(mpsr->nodes, mpsr->user_nodes,
 						sizeof(int) * count);
+			}
 			mpsr->nodes_ready = 1;
 			break;
 		case 1:
@@ -2292,8 +2297,10 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 			       sizeof(void *) * count);
 			break;
 		case 1:
+			if (mpsr->user_nodes) {
 				memcpy(mpsr->nodes, mpsr->user_nodes,
 						sizeof(int) * count);
+			}
 			mpsr->nodes_ready = 1;
 			break;
 		case 2:
@@ -2322,8 +2329,10 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 			       sizeof(void *) * (count / 2));
 			break;
 		case 2:
+			if (mpsr->user_nodes) {
 				memcpy(mpsr->nodes, mpsr->user_nodes,
 						sizeof(int) * count);
+			}
 			mpsr->nodes_ready = 1;
 			break;
 		case 3:
@@ -2349,6 +2358,7 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 	}
 
 	/* NUMA verification in parallel */
+	if (mpsr->user_nodes) {
 		for (i = i_s; i < i_e; i++) {
 			if (mpsr->nodes[i] < 0 ||
 					mpsr->nodes[i] >= ihk_mc_get_nr_numa_nodes() ||
@@ -2358,6 +2368,7 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 				break;
 			}
 		}
+	}
 
 	/* Barrier */
 	ihk_atomic_inc(&mpsr->phase_done);
@@ -2387,7 +2398,7 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 
 		/* PTE valid? */
 		if (!mpsr->ptep[i] || !pte_is_present(mpsr->ptep[i])) {
-			mpsr->status[i] = -ENOENT;
+			mpsr->status[i] = -EFAULT;
 			mpsr->ptep[i] = NULL;
 			continue;
 		}
@@ -2451,6 +2462,26 @@ pte_out:
 	dkprintf("%s: phase %d done\n", __FUNCTION__, phase);
 	++phase;
 
+	/*
+	 * When nodes array is NULL, move_pages doesn't move any pages,
+	 * instead will return the node where each page
+	 * currently resides by status array.
+	 */
+	if (!mpsr->user_nodes) {
+		/* get nid in parallel */
+		for (i = i_s; i < i_e; i++) {
+			if (mpsr->status[i] < 0) {
+				continue;
+			}
+			mpsr->status[i] = phys_to_nid(
+					pte_get_phys(mpsr->ptep[i]));
+		}
+		mpsr->phase_ret = 0;
+		goto out;  // return node information
+	}
+
+	/* Processing of move pages */
+
 	if (cpu_index == 0) {
 		/* Allocate new pages on target NUMA nodes */
 		for (i = 0; i < count; i++) {
@@ -2463,8 +2494,11 @@ pte_out:
 
 			/* TODO: store pgalign info in an array as well? */
 			if (mpsr->nr_pages[i] > 1) {
-				if (mpsr->nr_pages[i] * PAGE_SIZE == PTL2_SIZE)
+				int nr_pages;
-					pgalign = PTL2_SHIFT - PTL1_SHIFT;
+
+				for (pgalign = 0, nr_pages = mpsr->nr_pages[i];
+				     nr_pages != 1; pgalign++, nr_pages >>= 1) {
+				}
 			}
 
 			dst = ihk_mc_alloc_aligned_pages_node(mpsr->nr_pages[i],

arch/x86_64/kernel/syscall.c

@@ -32,6 +32,7 @@
 #include <limits.h>
 #include <syscall.h>
 #include <rusage_private.h>
+#include <memory.h>
 #include <ihk/debug.h>
 
 void terminate_mcexec(int, int);
@@ -2302,8 +2303,10 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 			case 0:
 				memcpy(mpsr->virt_addr, mpsr->user_virt_addr,
 						sizeof(void *) * count);
+				if (mpsr->user_nodes) {
 					memcpy(mpsr->nodes, mpsr->user_nodes,
 							sizeof(int) * count);
+				}
 				memset(mpsr->ptep, 0, sizeof(pte_t) * count);
 				memset(mpsr->status, 0, sizeof(int) * count);
 				memset(mpsr->nr_pages, 0, sizeof(int) * count);
@@ -2321,8 +2324,10 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 			case 0:
 				memcpy(mpsr->virt_addr, mpsr->user_virt_addr,
 						sizeof(void *) * count);
+				if (mpsr->user_nodes) {
 					memcpy(mpsr->nodes, mpsr->user_nodes,
 							sizeof(int) * count);
+				}
 				mpsr->nodes_ready = 1;
 				break;
 			case 1:
@@ -2344,8 +2349,10 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 						sizeof(void *) * count);
 				break;
 			case 1:
+				if (mpsr->user_nodes) {
 					memcpy(mpsr->nodes, mpsr->user_nodes,
 							sizeof(int) * count);
+				}
 				mpsr->nodes_ready = 1;
 				break;
 			case 2:
@@ -2374,8 +2381,10 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 						sizeof(void *) * (count / 2));
 				break;
 			case 2:
+				if (mpsr->user_nodes) {
 					memcpy(mpsr->nodes, mpsr->user_nodes,
 							sizeof(int) * count);
+				}
 				mpsr->nodes_ready = 1;
 				break;
 			case 3:
@@ -2401,6 +2410,7 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 	}
 
 	/* NUMA verification in parallel */
+	if (mpsr->user_nodes) {
 		for (i = i_s; i < i_e; i++) {
 			if (mpsr->nodes[i] < 0 ||
 					mpsr->nodes[i] >= ihk_mc_get_nr_numa_nodes() ||
@@ -2410,6 +2420,7 @@ int move_pages_smp_handler(int cpu_index, int nr_cpus, void *arg)
 				break;
 			}
 		}
+	}
 
 	/* Barrier */
 	ihk_atomic_inc(&mpsr->phase_done);
@@ -2503,6 +2514,26 @@ pte_out:
 	dkprintf("%s: phase %d done\n", __FUNCTION__, phase);
 	++phase;
 
+	/*
+	 * When nodes array is NULL, move_pages doesn't move any pages,
+	 * instead will return the node where each page
+	 * currently resides by status array.
+	 */
+	if (!mpsr->user_nodes) {
+		/* get nid in parallel */
+		for (i = i_s; i < i_e; i++) {
+			if (mpsr->status[i] < 0) {
+				continue;
+			}
+			mpsr->status[i] = phys_to_nid(
+					pte_get_phys(mpsr->ptep[i]));
+		}
+		mpsr->phase_ret = 0;
+		goto out;  // return node information
+	}
+
+	/* Processing of move pages */
+
 	if (cpu_index == 0) {
 		/* Allocate new pages on target NUMA nodes */
 		for (i = 0; i < count; i++) {

kernel/syscall.c

@@ -10250,7 +10250,7 @@ SYSCALL_DECLARE(move_pages)
 	struct move_pages_smp_req mpsr;
 
 	struct process_vm *vm = cpu_local_var(current)->vm;
-	int ret = 0;
+	int i, ret = 0;
 
 	unsigned long t_s, t_e;
 
@@ -10260,18 +10260,20 @@ SYSCALL_DECLARE(move_pages)
 	if (pid) {
 		kprintf("%s: ERROR: only self (pid == 0)"
 				" is supported\n", __FUNCTION__);
-		return -EINVAL;
+		ret = -EINVAL;
+		goto out;
 	}
 
-	switch (flags) {
+    /* Check flags */
-	case MPOL_MF_MOVE_ALL:
+	if (flags & ~(MPOL_MF_MOVE|MPOL_MF_MOVE_ALL)) {
+		ret = -EINVAL;
+		goto out;
+	}
+	if (flags & MPOL_MF_MOVE_ALL) {
 		kprintf("%s: ERROR: MPOL_MF_MOVE_ALL"
 			" not supported\n", __func__);
-		return -EINVAL;
+		ret = -EINVAL;
-	case MPOL_MF_MOVE:
+		goto out;
-		break;
-	default:
-		return -EINVAL;
 	}
 
 	/* Allocate kernel arrays */
@@ -10319,7 +10321,7 @@ t_e = rdtsc(); kprintf("%s: init malloc: %lu \n", __FUNCTION__, t_e - t_s); t_s
 		goto dealloc_out;
 	}
 
-	if (verify_process_vm(cpu_local_var(current)->vm,
+	if (user_nodes && verify_process_vm(cpu_local_var(current)->vm,
 				user_nodes, sizeof(int) * count)) {
 		ret = -EFAULT;
 		goto dealloc_out;
@@ -10330,6 +10332,18 @@ t_e = rdtsc(); kprintf("%s: init malloc: %lu \n", __FUNCTION__, t_e - t_s); t_s
 		ret = -EFAULT;
 		goto dealloc_out;
 	}
+
+	/* Check node ID */
+	if (user_nodes) {
+		copy_from_user(nodes, user_nodes, sizeof(int) * count);
+		for (i = 0; i < count; i++) {
+			if (nodes[i] < 0 || nodes[i] >= ihk_mc_get_nr_numa_nodes()) {
+				ret = -ENODEV;
+				goto dealloc_out;
+			}
+		}
+	}
+
 t_e = rdtsc(); kprintf("%s: init verify: %lu \n", __FUNCTION__, t_e - t_s); t_s = t_e;
 
 #if 0
@@ -10422,6 +10436,7 @@ dealloc_out:
 	kfree(ptep);
 	kfree(dst_phys);
 
+out:
 	return ret;
 }
 

test/issues/1523/C1523.sh

@@ -0,0 +1,30 @@
+#/bin/sh
+
+USELTP=1
+USEOSTEST=0
+
+MCREBOOT=0
+. ../../common.sh
+
+BOOTPARAM="${BOOTPARAM} -e anon_on_demand"
+mcreboot
+
+issue="1523"
+tid=01
+
+for tp in move_pages01 move_pages02 move_pages04 move_pages06 move_pages09 move_pages10
+do
+	tname=`printf "C${issue}T%02d" ${tid}`
+	echo "*** ${tname} start *******************************"
+	sudo $MCEXEC $LTPBIN/$tp 2>&1 | tee $tp.txt
+	ok=`grep PASS $tp.txt | wc -l`
+	ng=`grep FAIL $tp.txt | wc -l`
+	if [ $ng = 0 ]; then
+		echo "*** ${tname} PASSED ($ok)"
+	else
+		echo "*** ${tname} FAILED (ok=$ok ng=$ng)"
+	fi
+	let tid++
+	echo ""
+done
+

test/issues/1523/Makefile

@@ -0,0 +1,11 @@
+CFLAGS=-g
+LDFLAGS=
+
+TARGET=
+
+all: $(TARGET)
+
+test: all
+	./C1523.sh
+clean:
+	rm -f $(TARGET) *.o *.txt

test/issues/1523/README

@@ -0,0 +1,21 @@
+【Issue#1523 動作確認】
+□ テスト内容
+1. 以下のLTPがPASSすることを確認する
+  - move_pages01
+  - move_pages02
+  - move_pages04
+  - move_pages06
+  - move_pages09
+  - move_pages10
+
+□ 実行手順
+$ make test
+
+McKernelのインストール先や、OSTEST, LTPの配置場所は、
+$HOME/.mck_test_config を参照している
+.mck_test_config は、McKernelをビルドした際に生成されるmck_test_config.sample ファイルを
+$HOMEにコピーし、適宜編集する
+
+□ 実行結果
+x86_64_result.log aarch64_result.log 参照。
+すべての項目をPASSしていることを確認。

test/issues/1523/aarch64_result.log

@@ -0,0 +1,25 @@
+mcstop+release.sh ... done
+mcreboot.sh -c 37-43,49-55 -m 2G@2,2G@3 -r 37-43:36+49-55:48 -O -e anon_on_demand ... done
+*** C1523T01 start *******************************
+move_pages01    1  TPASS  :  pages are present in expected nodes
+*** C1523T01 PASSED (1)
+
+*** C1523T02 start *******************************
+move_pages02    1  TPASS  :  pages are present in expected nodes
+*** C1523T02 PASSED (1)
+
+*** C1523T03 start *******************************
+move_pages04    1  TPASS  :  status[1] has expected value
+*** C1523T03 PASSED (1)
+
+*** C1523T04 start *******************************
+move_pages06    1  TPASS  :  move_pages failed with ENODEV as expected
+*** C1523T04 PASSED (1)
+
+*** C1523T05 start *******************************
+move_pages09    1  TPASS  :  move_pages succeeded
+*** C1523T05 PASSED (1)
+
+*** C1523T06 start *******************************
+move_pages10    1  TPASS  :  move_pages failed with EINVAL as expected
+*** C1523T06 PASSED (1)

test/issues/1523/x86_64_result.log

@@ -0,0 +1,25 @@
+mcstop+release.sh ... done
+mcreboot.sh -c 1-7,9-15,17-23,25-31 -m 10G@0,10G@1 -r 1-7:0+9-15:8+17-23:16+25-31:24 -O -e anon_on_demand ... done
+*** C1523T01 start *******************************
+move_pages01    1  TPASS  :  pages are present in expected nodes
+*** C1523T01 PASSED (1)
+
+*** C1523T02 start *******************************
+move_pages02    1  TPASS  :  pages are present in expected nodes
+*** C1523T02 PASSED (1)
+
+*** C1523T03 start *******************************
+move_pages04    1  TPASS  :  status[1] has expected value
+*** C1523T03 PASSED (1)
+
+*** C1523T04 start *******************************
+move_pages06    1  TPASS  :  move_pages failed with ENODEV as expected
+*** C1523T04 PASSED (1)
+
+*** C1523T05 start *******************************
+move_pages09    1  TPASS  :  move_pages succeeded
+*** C1523T05 PASSED (1)
+
+*** C1523T06 start *******************************
+move_pages10    1  TPASS  :  move_pages failed with EINVAL as expected
+*** C1523T06 PASSED (1)