support vdso which borrows clocksource from linux

executer/include/uprotocol.h

@@ -87,6 +87,8 @@ struct program_load_desc {
 	int pgid;
 	int cred[8];
 	int reloc;
+	char enable_vdso;
+	char padding[7];
 	unsigned long entry;
 	unsigned long user_start;
 	unsigned long user_end;

executer/kernel/mcctrl/arch/x86_64/archdeps.c

@@ -2,6 +2,58 @@
 #include "../../config.h"
 #include "../../mcctrl.h"
 
+#ifdef MCCTRL_KSYM_vdso_image_64
+#if MCCTRL_KSYM_vdso_image_64
+struct vdso_image *vdso_image = (void *)MCCTRL_KSYM_vdso_image_64;
+#endif
+#endif
+
+#ifdef MCCTRL_KSYM_vdso_start
+#if MCCTRL_KSYM_vdso_start
+void *vdso_start = (void *)MCCTRL_KSYM_vdso_start;
+#endif
+#endif
+
+#ifdef MCCTRL_KSYM_vdso_end
+#if MCCTRL_KSYM_vdso_end
+void *vdso_end = (void *)MCCTRL_KSYM_vdso_end;
+#endif
+#endif
+
+#ifdef MCCTRL_KSYM_vdso_pages
+#if MCCTRL_KSYM_vdso_pages
+struct page **vdso_pages = (void *)MCCTRL_KSYM_vdso_pages;
+#endif
+#endif
+
+#ifdef MCCTRL_KSYM___vvar_page
+#if MCCTRL_KSYM___vvar_page
+void *__vvar_page = (void *)MCCTRL_KSYM___vvar_page;
+#endif
+#endif
+
+long *hpet_addressp
+#ifdef MCCTRL_KSYM_hpet_address
+#if MCCTRL_KSYM_hpet_address
+	= (void *)MCCTRL_KSYM_hpet_address;
+#else
+	= &hpet_address;
+#endif
+#else
+	= NULL;
+#endif
+
+void **hv_clockp
+#ifdef MCCTRL_KSYM_hv_clock
+#if MCCTRL_KSYM_hv_clock
+	= (void *)MCCTRL_KSYM_hv_clock;
+#else
+	= &hv_clock;
+#endif
+#else
+	= NULL;
+#endif
+
 unsigned long
 reserve_user_space_common(struct mcctrl_usrdata *usrdata, unsigned long start, unsigned long end);
 
@@ -36,3 +88,107 @@ reserve_user_space(struct mcctrl_usrdata *usrdata, unsigned long *startp, unsign
 	*endp = end;
 	return 0;
 }
+
+void get_vdso_info(ihk_os_t os, long vdso_rpa)
+{
+	ihk_device_t dev = ihk_os_to_dev(os);
+	long vdso_pa;
+	struct vdso *vdso;
+	size_t size;
+	int i;
+
+	vdso_pa = ihk_device_map_memory(dev, vdso_rpa, sizeof(*vdso));
+	vdso = ihk_device_map_virtual(dev, vdso_pa, sizeof(*vdso), NULL, 0);
+
+	memset(vdso, 0, sizeof(*vdso));
+
+	/* VDSO pages */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(3,16,0)
+	size = vdso_image->size;
+	vdso->vdso_npages = size >> PAGE_SHIFT;
+
+	if (vdso->vdso_npages > VDSO_MAXPAGES) {
+		vdso->vdso_npages = 0;
+		goto out;
+	}
+
+	for (i = 0; i < vdso->vdso_npages; ++i) {
+		vdso->vdso_physlist[i] = virt_to_phys(
+				vdso_image->data + (i * PAGE_SIZE));
+	}
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
+	size = vdso_end - vdso_start;
+	size = (size + PAGE_SIZE - 1) & PAGE_MASK;
+
+	vdso->vdso_npages = size >> PAGE_SHIFT;
+	if (vdso->vdso_npages > VDSO_MAXPAGES) {
+		vdso->vdso_npages = 0;
+		goto out;
+	}
+
+	for (i = 0; i < vdso->vdso_npages; ++i) {
+		vdso->vdso_physlist[i] = page_to_phys(vdso_pages[i]);
+	}
+#endif
+
+	/* VVAR page */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4,5,0)
+	vdso->vvar_is_global = 0;
+	vdso->vvar_virt = -3 * PAGE_SIZE;
+	vdso->vvar_phys = virt_to_phys(__vvar_page);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3,17,0)
+	vdso->vvar_is_global = 0;
+	vdso->vvar_virt = -2 * PAGE_SIZE;
+	vdso->vvar_phys = virt_to_phys(__vvar_page);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3,16,0)
+	vdso->vvar_is_global = 0;
+	vdso->vvar_virt = (void *)(vdso->vdso_npages * PAGE_SIZE);
+	vdso->vvar_phys = virt_to_phys(__vvar_page);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3,1,0)
+	vdso->vvar_is_global = 1;
+	vdso->vvar_virt = (void *)fix_to_virt(VVAR_PAGE);
+	vdso->vvar_phys = virt_to_phys(__vvar_page);
+#endif
+
+	/* HPET page */
+	if (hpet_addressp && *hpet_addressp) {
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4,5,0)
+		vdso->hpet_is_global = 0;
+		vdso->hpet_virt = -2 * PAGE_SIZE;
+		vdso->hpet_phys = *hpet_addressp;
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3,17,0)
+		vdso->hpet_is_global = 0;
+		vdso->hpet_virt = -1 * PAGE_SIZE;
+		vdso->hpet_phys = *hpet_addressp;
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3,16,0)
+		vdso->hpet_is_global = 0;
+		vdso->hpet_virt = (void *)((vdso->vdso_npages + 1) * PAGE_SIZE);
+		vdso->hpet_phys = *hpet_addressp;
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,23)
+		vdso->hpet_is_global = 1;
+		vdso->hpet_virt = (void *)fix_to_virt(VSYSCALL_HPET);
+		vdso->hpet_phys = *hpet_addressp;
+#endif
+	}
+
+	/* struct pvlock_vcpu_time_info table */
+	if (hv_clockp && *hv_clockp) {
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(4,5,0)
+		vdso->pvti_is_global = 0;
+		vdso->pvti_virt = -1 * PAGE_SIZE;
+		vdso->pvti_phys = virt_to_phys(*hv_clockp);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(3,8,0)
+		vdso->pvti_is_global = 1;
+		vdso->pvti_virt = (void *)fix_to_virt(PVCLOCK_FIXMAP_BEGIN);
+		vdso->pvti_phys = virt_to_phys(*hv_clockp);
+#endif
+	}
+
+out:
+	wmb();
+	vdso->busy = 0;
+
+	ihk_device_unmap_virtual(dev, vdso, sizeof(*vdso));
+	ihk_device_unmap_memory(dev, vdso_pa, sizeof(*vdso));
+	return;
+} /* get_vdso_info() */

executer/kernel/mcctrl/ikc.c

@@ -95,6 +95,10 @@ static int syscall_packet_handler(struct ihk_ikc_channel_desc *c,
 		delete_tid_entry(ihk_host_os_get_index(__os), pisp->pid, pisp->arg);
 		break;
 
+	case SCD_MSG_GET_VDSO_INFO:
+		get_vdso_info(__os, pisp->arg);
+		break;
+
 	default:
 		printk(KERN_ERR "mcctrl:syscall_packet_handler:"
 				"unknown message (%d.%d.%d.%d.%d.%#lx)\n",

executer/kernel/mcctrl/mcctrl.h

@@ -55,6 +55,7 @@
 #define SCD_MSG_SYSCALL_ONESIDE         0x4
 #define SCD_MSG_SEND_SIGNAL     	0x8
 #define SCD_MSG_CLEANUP_PROCESS         0x9
+#define SCD_MSG_GET_VDSO_INFO           0xa
 
 #define	SCD_MSG_PROCFS_CREATE		0x10
 #define	SCD_MSG_PROCFS_DELETE		0x11
@@ -230,8 +231,6 @@ struct mcctrl_signal {
 int mcctrl_ikc_send(ihk_os_t os, int cpu, struct ikc_scd_packet *pisp);
 int mcctrl_ikc_send_msg(ihk_os_t os, int cpu, int msg, int ref, unsigned long arg);
 int mcctrl_ikc_is_valid_thread(ihk_os_t os, int cpu);
-int reserve_user_space(struct mcctrl_usrdata *usrdata, unsigned long *startp,
-		unsigned long *endp);
 
 ihk_os_t osnum_to_os(int n);
 
@@ -275,4 +274,26 @@ void procfs_exit(int osnum);
 /* sysfs_files.c */
 void setup_sysfs_files(ihk_os_t os);
 
+/* archdep.c */
+#define VDSO_MAXPAGES 2
+struct vdso {
+	long busy;
+	int vdso_npages;
+	char vvar_is_global;
+	char hpet_is_global;
+	char pvti_is_global;
+	char padding;
+	long vdso_physlist[VDSO_MAXPAGES];
+	void *vvar_virt;
+	long vvar_phys;
+	void *hpet_virt;
+	long hpet_phys;
+	void *pvti_virt;
+	long pvti_phys;
+};
+
+int reserve_user_space(struct mcctrl_usrdata *usrdata, unsigned long *startp,
+		unsigned long *endp);
+void get_vdso_info(ihk_os_t os, long vdso_pa);
+
 #endif

executer/user/mcexec.c

@@ -63,6 +63,7 @@
 #include <include/generated/uapi/linux/version.h>
 #include <sys/user.h>
 #include "../include/uprotocol.h"
+#include <getopt.h>
 
 //#define DEBUG
 
@@ -132,6 +133,7 @@ static char *exec_path = NULL;
 static char *altroot;
 static const char rlimit_stack_envname[] = "MCKERNEL_RLIMIT_STACK";
 static int ischild;
+static int enable_vdso = 1;
 
 struct fork_sync {
 	pid_t pid;
@@ -1231,6 +1233,24 @@ static int rlimits[] = {
 
 char dev[64];
 
+static struct option mcexec_options[] = {
+	{
+		.name =		"disable-vdso",
+		.has_arg =	no_argument,
+		.flag =		&enable_vdso,
+		.val =		0,
+	},
+	{
+		.name =		"enable-vdso",
+		.has_arg =	no_argument,
+		.flag =		&enable_vdso,
+		.val =		1,
+	},
+
+	/* end */
+	{ NULL, 0, NULL, 0, },
+};
+
 int main(int argc, char **argv)
 {
 //	int fd;
@@ -1282,12 +1302,15 @@ int main(int argc, char **argv)
 	}
            
 	/* Parse options ("+" denotes stop at the first non-option) */
-	while ((opt = getopt(argc, argv, "+c:")) != -1) {
+	while ((opt = getopt_long(argc, argv, "+c:", mcexec_options, NULL)) != -1) {
 		switch (opt) {
 			case 'c':
 				target_core = atoi(optarg);
 				break;
 			
+			case 0:	/* long opt */
+				break;
+
 			default: /* '?' */
 				print_usage(argv);
 				exit(EXIT_FAILURE);
@@ -1416,6 +1439,8 @@ int main(int argc, char **argv)
 	//print_flat(args);
 
 	desc->cpu = target_core;
+	desc->enable_vdso = enable_vdso;
+
 	p = getenv(rlimit_stack_envname);
 	if (p) {
 		errno = 0;