The build is successful, testing is required

Change-Id: I503233b393e5bfec003d407512d1028de3a60946

.gitmodules

@@ -1,12 +1,12 @@
 [submodule "ihk"]
 	path = ihk
-	url = https://github.com/RIKEN-SysSoft/ihk.git
+	url = https://github.com/ihkmckernel/ihk.git
 [submodule "executer/user/lib/libdwarf/libdwarf"]
 	path = executer/user/lib/libdwarf/libdwarf
 	url = https://github.com/bgerofi/libdwarf.git
 [submodule "executer/user/lib/syscall_intercept"]
 	path = executer/user/lib/syscall_intercept
-	url = https://github.com/RIKEN-SysSoft/syscall_intercept.git
+	url = https://github.com/ihkmckernel/syscall_intercept.git
 [submodule "executer/user/lib/uti"]
 	path = executer/user/lib/uti
-	url = https://github.com/RIKEN-SysSoft/uti.git
+	url = https://github.com/ihkmckernel/uti.git

CMakeLists.txt

@@ -7,7 +7,7 @@ endif (NOT CMAKE_BUILD_TYPE)
 enable_language(C ASM)
 
 project(mckernel C ASM)
-set(MCKERNEL_VERSION "1.7.9")
+set(MCKERNEL_VERSION "1.8.0")
 
 # See "Fedora Packaging Guidelines -- Versioning"
 set(MCKERNEL_RELEASE "")
@@ -84,7 +84,8 @@ if(ENABLE_FUGAKU_HACKS)
 endif()
 
 # Fujitsu MPI tries to xpmem-attach segment with size of range size + 1?
-set(FJMPI_VERSION_COMMAND "a=\$(which mpifcc); b=\${a%/*/*}; c=\${b##*/}; d=\${c#*-}; echo \$d")
+#set(FJMPI_VERSION_COMMAND "a=\$(which mpifcc); b=\${a%/*/*}; c=\${b##*/}; d=\${c#*-}; echo \$d")
+set(FJMPI_VERSION_COMMAND "a=\$(which mpifort); b=\${a%/*/*}; c=\${b##*/}; d=\${c#*-}; echo \$d")
 execute_process(COMMAND bash -c "${FJMPI_VERSION_COMMAND}"
   OUTPUT_VARIABLE FJMPI_VERSION OUTPUT_STRIP_TRAILING_WHITESPACE)
 message("FJMPI_VERSION: ${FJMPI_VERSION}")
@@ -191,6 +192,7 @@ file(REMOVE_RECURSE ${tmpdir})
 file(MAKE_DIRECTORY ${tmpdir})
 file(WRITE ${tmpdir}/driver.c "#include <linux/module.h>\n")
 file(APPEND ${tmpdir}/driver.c "unsigned long MAP_KERNEL_START = MODULES_END - (1UL << 23);\n")
+file(APPEND ${tmpdir}/driver.c "MODULE_LICENSE(\"GPL\");\n")
 file(WRITE ${tmpdir}/Makefile "obj-m := driver.o\n")
 file(APPEND ${tmpdir}/Makefile "all:\n")
 file(APPEND ${tmpdir}/Makefile "\tmake ${KBUILD_MAKE_FLAGS_STR} -C ${KERNEL_DIR} M=${tmpdir} modules\n")
@@ -276,11 +278,11 @@ set(CMAKE_INSTALL_RPATH ${CMAKE_INSTALL_FULL_LIBDIR})
 
 # ihk: ultimately should support extrnal build, but add as subproject for now
 if (EXISTS ${PROJECT_SOURCE_DIR}/ihk/CMakeLists.txt)
-	set(IHK_SOURCE_DIR "ihk" CACHE STRINGS "path to ihk source directory from mckernel sources")
+	set(IHK_SOURCE_DIR "ihk" CACHE STRING "path to ihk source directory from mckernel sources")
 elseif (EXISTS ${PROJECT_SOURCE_DIR}/../ihk/CMakeLists.txt)
-	set(IHK_SOURCE_DIR "../ihk" CACHE STRINGS "path to ihk source directory from mckernel sources")
+	set(IHK_SOURCE_DIR "../ihk" CACHE STRING "path to ihk source directory from mckernel sources")
 else()
-	set(IHK_SOURCE_DIR "ihk" CACHE STRINGS "path to ihk source directory from mckernel sources")
+	set(IHK_SOURCE_DIR "ihk" CACHE STRING "path to ihk source directory from mckernel sources")
 endif()
 if (EXISTS ${PROJECT_SOURCE_DIR}/${IHK_SOURCE_DIR}/CMakeLists.txt)
 	set(IHK_FULL_SOURCE_DIR ${PROJECT_SOURCE_DIR}/${IHK_SOURCE_DIR})

arch/arm64/kernel/vdso.c

@@ -33,7 +33,6 @@ struct vdso {
 	long offset_sigtramp;
 };
 
-extern char vdso_start, vdso_end;
 static struct vdso vdso;
 
 struct tod_data_s tod_data

arch/x86_64/kernel/include/arch-memory.h

@@ -183,7 +183,7 @@ enum ihk_mc_pt_attribute {
 	PTATTR_WRITE_COMBINED = 0x40000,
 };
 
-enum ihk_mc_pt_attribute attr_mask;
+extern enum ihk_mc_pt_attribute attr_mask;
 
 static inline int pfn_is_write_combined(uintptr_t pfn)
 {

docs/NEWS.rst

@@ -1,3 +1,52 @@
+=============================================
+Version 1.8.0 (Mar 23, 2021)
+=============================================
+
+----------------------
+IHK major updates
+----------------------
+N/A
+
+------------------------
+IHK major bug fixes
+------------------------
+N/A
+
+----------------------
+McKernel major updates
+----------------------
+N/A
+
+------------------------
+McKernel major bug fixes
+------------------------
+#. profile: fix infinite recursion for allocation miss event
+#. Fugaku: MAP_LOCKED and pre-populate PMIx shared memory PFNs
+
+=============================================
+Version 1.7.10 (Mar 18, 2021)
+=============================================
+
+----------------------
+IHK major updates
+----------------------
+N/A
+
+------------------------
+IHK major bug fixes
+------------------------
+#. __ihk_device_detect_hungup: detect hungup via device-ioctl
+
+----------------------
+McKernel major updates
+----------------------
+N/A
+
+------------------------
+McKernel major bug fixes
+------------------------
+N/A
+
 =============================================
 Version 1.7.9 (Mar 17, 2021)
 =============================================

docs/users.rst

@@ -242,3 +242,5 @@ Limitations
 #. procfs entry creation done by Linux work queue could starve when
    Linux CPUs are flooded with system call offloads. LTP-2019
    sendmsg02 causes this issue.
+
+#. Linux kernel dump file doesn't include the memory allocated to McKernel. This is because of the issues in the implementation of the panic notifier handler.

executer/kernel/mcctrl/arch/arm64/archdeps.c

@@ -27,9 +27,12 @@
 #define D(fmt, ...) printk("%s(%d) " fmt, __func__, __LINE__, ##__VA_ARGS__)
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 0, 0)
-void *vdso_start;
-void *vdso_end;
-static struct vm_special_mapping (*vdso_spec)[2];
+void *mcctrl_vdso_start;
+void *mcctrl_vdso_end;
+static struct vm_special_mapping *mcctrl_vdso_spec;
+#if defined(RHEL_RELEASE_CODE) && RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(8, 4)
+static struct vdso_data **mcctrl_vdso_data;
+#endif
 #endif
 
 #ifdef ENABLE_TOFU
@@ -68,17 +71,36 @@ void __mcctrl_tof_utofu_mn_invalidate_range_end(
 int arch_symbols_init(void)
 {
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 0, 0)
-	vdso_start = (void *) kallsyms_lookup_name("vdso_start");
-	if (WARN_ON(!vdso_start))
+	mcctrl_vdso_start = (void *) kallsyms_lookup_name("vdso_start");
+	if (WARN_ON(!mcctrl_vdso_start))
 		return -EFAULT;
 
-	vdso_end = (void *) kallsyms_lookup_name("vdso_end");
-	if (WARN_ON(!vdso_end))
+	mcctrl_vdso_end = (void *) kallsyms_lookup_name("vdso_end");
+	if (WARN_ON(!mcctrl_vdso_end))
 		return -EFAULT;
 
-	vdso_spec = (void *) kallsyms_lookup_name("vdso_spec");
-	if (WARN_ON(!vdso_spec))
+#if defined(RHEL_RELEASE_CODE) && RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(8, 4)
+	mcctrl_vdso_spec = (void *) kallsyms_lookup_name("aarch64_vdso_maps");
+#else
+	mcctrl_vdso_spec = (void *) kallsyms_lookup_name("vdso_spec");
+#endif
+	if (WARN_ON(!mcctrl_vdso_spec))
 		return -EFAULT;
+
+	if (WARN_ON(!mcctrl_vdso_spec[0].name ||
+		strcmp(mcctrl_vdso_spec[0].name, "[vvar]")))
+		return -EFAULT;
+
+	if (WARN_ON(!mcctrl_vdso_spec[1].name ||
+		strcmp(mcctrl_vdso_spec[1].name, "[vdso]")))
+		return -EFAULT;
+
+#if defined(RHEL_RELEASE_CODE) && RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(8, 4)
+	mcctrl_vdso_data = (struct vdso_data **) kallsyms_lookup_name("vdso_data");
+	if (WARN_ON(!mcctrl_vdso_data || !*mcctrl_vdso_data))
+		return -EFAULT;
+#endif
+
 #endif
 
 #ifdef ENABLE_TOFU
@@ -212,7 +234,7 @@ static long elf_search_vdso_sigtramp(void)
 	Elf64_Sym *sym = NULL;
 
 	/* ELF header */
-	eh = (Elf64_Ehdr *)vdso_start;
+	eh = (Elf64_Ehdr *)mcctrl_vdso_start;
 	if (eh == NULL) {
 		D("vdso_start is NULL.\n");
 		goto out;
@@ -233,8 +255,8 @@ static long elf_search_vdso_sigtramp(void)
 	/* Search dynsym-table and dynstr-table offset
 	 * from section header table
 	 */
-	tmp_sh = (Elf64_Shdr *)(vdso_start + eh->e_shoff);
-	shstr = vdso_start + (tmp_sh + eh->e_shstrndx)->sh_offset;
+	tmp_sh = (Elf64_Shdr *)(mcctrl_vdso_start + eh->e_shoff);
+	shstr = mcctrl_vdso_start + (tmp_sh + eh->e_shstrndx)->sh_offset;
 	for (i = 0; i < eh->e_shnum; i++, tmp_sh++) {
 		if (tmp_sh->sh_type == SHT_DYNSYM) {
 			sym_sh = tmp_sh;
@@ -242,7 +264,7 @@ static long elf_search_vdso_sigtramp(void)
 
 		if (tmp_sh->sh_type == SHT_STRTAB &&
 		    !strcmp(&shstr[tmp_sh->sh_name], ".dynstr")) {
-			dynstr = vdso_start + tmp_sh->sh_offset;
+			dynstr = mcctrl_vdso_start + tmp_sh->sh_offset;
 		}
 	}
 
@@ -257,7 +279,7 @@ static long elf_search_vdso_sigtramp(void)
 	}
 
 	/* Search __kernel_rt_sigreturn offset from dynsym-table */
-	sym = (Elf64_Sym *)(vdso_start + sym_sh->sh_offset);
+	sym = (Elf64_Sym *)(mcctrl_vdso_start + sym_sh->sh_offset);
 	for (i = 0; (i * sym_sh->sh_entsize) < sym_sh->sh_size; i++, sym++) {
 		if (!strcmp(dynstr + sym->st_name, "__kernel_rt_sigreturn")) {
 			ans = sym->st_value;
@@ -282,9 +304,9 @@ void get_vdso_info(ihk_os_t os, long vdso_rpa)
 	vdso = ihk_device_map_virtual(dev, vdso_pa, sizeof(*vdso), NULL, 0);
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4,0,0)
-	vvar_map = &(*vdso_spec)[0];
-	vdso_map = &(*vdso_spec)[1];
-	nr_vdso_page = ((vdso_end - vdso_start) + PAGE_SIZE - 1) >> PAGE_SHIFT;
+	vvar_map = &mcctrl_vdso_spec[0];
+	vdso_map = &mcctrl_vdso_spec[1];
+	nr_vdso_page = ((mcctrl_vdso_end - mcctrl_vdso_start) + PAGE_SIZE - 1) >> PAGE_SHIFT;
 
 	/* VDSO pages */
 	//D("nr_vdso_page:%d\n", nr_vdso_page);
@@ -298,7 +320,11 @@ void get_vdso_info(ihk_os_t os, long vdso_rpa)
 
 	/* VVAR page */
 	//D("vdso->vvar_phys:0x#lx\n", vdso->vvar_phys);
+#if defined(RHEL_RELEASE_CODE) && RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(8, 4)
+	vdso->vvar_phys = __pfn_to_phys(sym_to_pfn(*mcctrl_vdso_data));
+#else
 	vdso->vvar_phys = page_to_phys(*vvar_map->pages);
+#endif
 
 	/* offsets */
 	vdso->lbase = VDSO_LBASE;

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

@@ -6,8 +6,13 @@
 #include <asm/vgtod.h>
 #include "config.h"
 #include "../../mcctrl.h"
+#include "../../kallsyms_compat.h"
 
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5,8,0) && defined(CONFIG_X86_VSYSCALL_EMULATION)
 #define gtod (&VVAR(vsyscall_gtod_data))
+#else
+#define gtod NULL
+#endif
 
 //#define SC_DEBUG
 
@@ -24,36 +29,36 @@ static void *vdso_start;
 static void *vdso_end;
 static struct page **vdso_pages;
 #endif
-static void *__vvar_page;
+static void *__vvar_page_ptr;
 static long *hpet_address;
 static void **hv_clock;
 
 int arch_symbols_init(void)
 {
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3, 16, 0)
-	_vdso_image_64 = (void *) kallsyms_lookup_name("vdso_image_64");
+	_vdso_image_64 = (void *) mcctrl_lookup_name("vdso_image_64");
 	if (WARN_ON(!_vdso_image_64))
 		return -EFAULT;
 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(2, 6, 23)
-	vdso_start = (void *) kallsyms_lookup_name("vdso_start");
+	vdso_start = (void *) mcctrl_lookup_name("vdso_start");
 	if (WARN_ON(!vdso_start))
 		return -EFAULT;
 
-	vdso_end = (void *) kallsyms_lookup_name("vdso_end");
+	vdso_end = (void *) mcctrl_lookup_name("vdso_end");
 	if (WARN_ON(!vdso_end))
 		return -EFAULT;
 
-	vdso_pages = (void *) kallsyms_lookup_name("vdso_pages");
+	vdso_pages = (void *) mcctrl_lookup_name("vdso_pages");
 	if (WARN_ON(!vdso_pages))
 		return -EFAULT;
 #endif
 
-	__vvar_page = (void *) kallsyms_lookup_name("__vvar_page");
-	if (WARN_ON(!__vvar_page))
+	__vvar_page_ptr = (void *) &__vvar_page;
+	if (WARN_ON(!__vvar_page_ptr))
 		return -EFAULT;
 
-	hpet_address = (void *) kallsyms_lookup_name("hpet_address");
-	hv_clock = (void *) kallsyms_lookup_name("hv_clock");
+	hpet_address = (void *) mcctrl_lookup_name("hpet_address");
+	hv_clock = (void *) mcctrl_lookup_name("hv_clock");
 	return 0;
 }
 
@@ -93,18 +98,18 @@ reserve_user_space(struct mcctrl_usrdata *usrdata, unsigned long *startp, unsign
 #define	DESIRED_USER_END	0x800000000000
 #define	GAP_FOR_MCEXEC		0x008000000000UL
 	end = DESIRED_USER_END;
-	down_write(&current->mm->mmap_sem);
+	mmap_write_lock(current->mm);
 	vma = find_vma(current->mm, 0);
 	if (vma) {
 		end = (vma->vm_start - GAP_FOR_MCEXEC) & ~(GAP_FOR_MCEXEC - 1);
 	}
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,5,0)
-	up_write(&current->mm->mmap_sem);
+	mmap_write_unlock(current->mm);
 #endif
 	start = reserve_user_space_common(usrdata, start, end);
 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,5,0)
-	up_write(&current->mm->mmap_sem);
+	mmap_write_unlock(current->mm);
 #endif
 
 	mutex_unlock(&usrdata->reserve_lock);
@@ -161,19 +166,19 @@ void get_vdso_info(ihk_os_t os, long vdso_rpa)
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4,5,0)
 	vdso->vvar_is_global = 0;
 	vdso->vvar_virt = (void *)(-3 * PAGE_SIZE);
-	vdso->vvar_phys = virt_to_phys(__vvar_page);
+	vdso->vvar_phys = virt_to_phys(__vvar_page_ptr);
 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(3,17,0)
 	vdso->vvar_is_global = 0;
 	vdso->vvar_virt = (void *)(-2 * PAGE_SIZE);
-	vdso->vvar_phys = virt_to_phys(__vvar_page);
+	vdso->vvar_phys = virt_to_phys(__vvar_page_ptr);
 #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);
+	vdso->vvar_phys = virt_to_phys(__vvar_page_ptr);
 #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);
+	vdso->vvar_phys = virt_to_phys(__vvar_page_ptr);
 #endif
 
 	/* HPET page */

executer/kernel/mcctrl/binfmt_mcexec.c

@@ -126,8 +126,8 @@ static int load_elf(struct linux_binprm *bprm
 			if(st == 0){
 				off = p & ~PAGE_MASK;
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4,10,0)
-				rc = get_user_pages_remote(current, bprm->mm,
-				                        bprm->p, 1, FOLL_FORCE, &page, NULL, NULL);
+				rc = get_user_pages_remote(bprm->mm,
+				                        bprm->p, 1, FOLL_FORCE, &page, NULL);
 #elif LINUX_VERSION_CODE >= KERNEL_VERSION(4,9,0)
 				rc = get_user_pages_remote(current, bprm->mm,
 				                        bprm->p, 1, FOLL_FORCE, &page, NULL);
@@ -234,7 +234,7 @@ static int load_elf(struct linux_binprm *bprm
 		kfree(pbuf);
 		return rc;
 	}
-	rc = copy_strings_kernel(1, &bprm->interp, bprm);
+	rc = copy_string_kernel(bprm->interp, bprm);
 	if (rc < 0){
 		fput(file);
 		kfree(pbuf);
@@ -242,7 +242,7 @@ static int load_elf(struct linux_binprm *bprm
 	}
 	bprm->argc++;
 	wp = MCEXEC_PATH;
-	rc = copy_strings_kernel(1, &wp, bprm);
+	rc = copy_string_kernel(wp, bprm);
 	if (rc){
 		fput(file);
 		kfree(pbuf);
@@ -260,19 +260,25 @@ static int load_elf(struct linux_binprm *bprm
 	fput(bprm->file);
 	bprm->file = file;
 
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5,8,0)
 	rc = prepare_binprm(bprm);
 	if (rc < 0){
 		kfree(pbuf);
 		return rc;
 	}
+#endif
 
 	kfree(pbuf);
 
+#if LINUX_VERSION_CODE < KERNEL_VERSION(5,8,0)
 	return search_binary_handler(bprm
 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,8,0)
 	                             , regs
 #endif
 );
+#else
+	return -ENOEXEC;
+#endif
 }
 
 static struct linux_binfmt mcexec_format = {

executer/kernel/mcctrl/control.c

@@ -978,7 +978,9 @@ static long mcexec_get_cpuset(ihk_os_t os, unsigned long arg)
 		node = linux_numa_2_mckernel_numa(udp,
 				cpu_to_node(mckernel_cpu_2_linux_cpu(udp, cpu_prev)));
 
-		for_each_cpu_not(cpu, cpus_used) {
+		for (cpu = 0; cpu < nr_cpu_ids; cpu++) {
+			if (cpumask_test_cpu(cpu, cpus_used))
+				continue;
 			/* Invalid CPU? */
 			if (cpu >= udp->cpu_info->n_cpus)
 				break;
@@ -1797,7 +1799,7 @@ out:
 }
 
 LIST_HEAD(mckernel_exec_files);
-DEFINE_SEMAPHORE(mckernel_exec_file_lock);
+DEFINE_SEMAPHORE(mckernel_exec_file_lock, 1);
  
 
 struct mckernel_exec_file {
@@ -3262,8 +3264,9 @@ mcexec_uti_attr(ihk_os_t os, struct uti_attr_desc __user *_desc)
 			            &lcache_topo->shared_cpu_map);
 		}
 		else {
-			cpumask_complement(wkmask,
-			                   &lcache_topo->shared_cpu_map);
+			bitmap_complement(cpumask_bits(wkmask),
+			                  cpumask_bits(&lcache_topo->shared_cpu_map),
+			                  nr_cpumask_bits);
 			cpumask_and(cpuset, cpuset, wkmask);
 		}
 	}
@@ -3276,8 +3279,9 @@ mcexec_uti_attr(ihk_os_t os, struct uti_attr_desc __user *_desc)
 			            &lcache_topo->shared_cpu_map);
 		}
 		else {
-			cpumask_complement(wkmask,
-			                   &lcache_topo->shared_cpu_map);
+			bitmap_complement(cpumask_bits(wkmask),
+			                  cpumask_bits(&lcache_topo->shared_cpu_map),
+			                  nr_cpumask_bits);
 			cpumask_and(cpuset, cpuset, wkmask);
 		}
 	}
@@ -3290,8 +3294,9 @@ mcexec_uti_attr(ihk_os_t os, struct uti_attr_desc __user *_desc)
 			            &lcache_topo->shared_cpu_map);
 		}
 		else {
-			cpumask_complement(wkmask,
-			                   &lcache_topo->shared_cpu_map);
+			bitmap_complement(cpumask_bits(wkmask),
+			                  cpumask_bits(&lcache_topo->shared_cpu_map),
+			                  nr_cpumask_bits);
 			cpumask_and(cpuset, cpuset, wkmask);
 		}
 	}

executer/kernel/mcctrl/driver.c

@@ -34,6 +34,7 @@
 #include <linux/version.h>
 #include "mcctrl.h"
 #include <ihk/ihk_host_user.h>
+#include "kallsyms_compat.h"
 
 #define OS_MAX_MINOR 64
 
@@ -237,75 +238,84 @@ struct inode_operations *mcctrl_hugetlbfs_inode_operations;
 
 static int symbols_init(void)
 {
+	int ret;
+	
+	/* Initialize kallsyms compatibility layer */
+	ret = init_kallsyms_lookup();
+	if (ret) {
+		pr_err("Failed to initialize kallsyms compatibility layer: %d\n", ret);
+		return ret;
+	}
+
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4,17,0)
-	mcctrl_sys_mount = (void *) kallsyms_lookup_name("ksys_mount");
+	mcctrl_sys_mount = (void *) mcctrl_lookup_name("ksys_mount");
 #else
-	mcctrl_sys_mount = (void *) kallsyms_lookup_name("sys_mount");
+	mcctrl_sys_mount = (void *) mcctrl_lookup_name("sys_mount");
 #if defined(CONFIG_X86_64_SMP)
 	if (!mcctrl_sys_mount)
 		mcctrl_sys_mount =
-			(void *) kallsyms_lookup_name("__x64_sys_mount");
+			(void *) mcctrl_lookup_name("__x64_sys_mount");
 #endif
 #endif
 	if (WARN_ON(!mcctrl_sys_mount))
 		return -EFAULT;
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4,17,0)
-	mcctrl_sys_umount = (void *) kallsyms_lookup_name("ksys_umount");
+	mcctrl_sys_umount = (void *) mcctrl_lookup_name("ksys_umount");
 #else
-	mcctrl_sys_umount = (void *) kallsyms_lookup_name("sys_umount");
+	mcctrl_sys_umount = (void *) mcctrl_lookup_name("sys_umount");
 #if defined(CONFIG_X86_64_SMP)
 	if (!mcctrl_sys_umount)
 		mcctrl_sys_umount =
-			(void *) kallsyms_lookup_name("__x64_sys_umount");
+			(void *) mcctrl_lookup_name("__x64_sys_umount");
 #endif
 #endif
 	if (WARN_ON(!mcctrl_sys_umount))
 		return -EFAULT;
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4,17,0)
-	mcctrl_sys_unshare = (void *) kallsyms_lookup_name("ksys_unshare");
+	mcctrl_sys_unshare = (void *) mcctrl_lookup_name("ksys_unshare");
 #else
-	mcctrl_sys_unshare = (void *) kallsyms_lookup_name("sys_unshare");
+	mcctrl_sys_unshare = (void *) mcctrl_lookup_name("sys_unshare");
 #if defined(CONFIG_X86_64_SMP)
 	if (!mcctrl_sys_unshare)
 		mcctrl_sys_unshare =
-			(void *) kallsyms_lookup_name("__x64_sys_unshare");
+			(void *) mcctrl_lookup_name("__x64_sys_unshare");
 #endif
 #endif
 	if (WARN_ON(!mcctrl_sys_unshare))
 		return -EFAULT;
 
 	mcctrl_sched_setaffinity =
-		(void *) kallsyms_lookup_name("sched_setaffinity");
+		(void *) mcctrl_lookup_name("sched_setaffinity");
 	if (WARN_ON(!mcctrl_sched_setaffinity))
 		return -EFAULT;
 
 	mcctrl_sched_setscheduler_nocheck =
-		(void *) kallsyms_lookup_name("sched_setscheduler_nocheck");
+		(void *) mcctrl_lookup_name("sched_setscheduler_nocheck");
 	if (WARN_ON(!mcctrl_sched_setscheduler_nocheck))
 		return -EFAULT;
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(4,17,0)
-	mcctrl_sys_readlinkat = (void *)kallsyms_lookup_name("do_readlinkat");
+	mcctrl_sys_readlinkat = (void *)mcctrl_lookup_name("do_readlinkat");
 #else
-	mcctrl_sys_readlinkat = (void *)kallsyms_lookup_name("sys_readlinkat");
+	mcctrl_sys_readlinkat = (void *)mcctrl_lookup_name("sys_readlinkat");
 #if defined(CONFIG_X86_64_SMP)
 	if (!mcctrl_sys_readlinkat)
 		mcctrl_sys_readlinkat =
-			(void *) kallsyms_lookup_name("__x64_sys_readlinkat");
+			(void *) mcctrl_lookup_name("__x64_sys_readlinkat");
 #endif
 #endif
 	if (WARN_ON(!mcctrl_sys_readlinkat))
 		return -EFAULT;
 
 	mcctrl_zap_page_range =
-		(void *) kallsyms_lookup_name("zap_page_range");
+		(void *) mcctrl_lookup_name("zap_page_range");
 	if (WARN_ON(!mcctrl_zap_page_range))
 		return -EFAULT;
 
 	mcctrl_hugetlbfs_inode_operations =
-		(void *) kallsyms_lookup_name("hugetlbfs_inode_operations");
+		(void *) mcctrl_lookup_name("hugetlbfs_inode_operations");
 	if (WARN_ON(!mcctrl_hugetlbfs_inode_operations))
 		return -EFAULT;
 
@@ -359,6 +369,7 @@ static void __exit mcctrl_exit(void)
 
 	binfmt_mcexec_exit();
 	uti_attr_finalize();
+	cleanup_kallsyms_lookup();
 #ifdef ENABLE_TOFU
 	mcctrl_tofu_restore_release_handlers();
 #endif

executer/kernel/mcctrl/futex.c

@@ -51,7 +51,7 @@ static long uti_wait_event(void *_resp, unsigned long nsec_timeout)
 	}
 }
 
-static int uti_clock_gettime(clockid_t clk_id, struct timespec *tp)
+static int uti_clock_gettime(clockid_t clk_id, struct timespec64 *tp)
 {
 	int ret = 0;
 	struct timespec64 ts64;
@@ -60,7 +60,7 @@ static int uti_clock_gettime(clockid_t clk_id, struct timespec *tp)
 			clk_id, CLOCK_REALTIME, CLOCK_MONOTONIC);
 	switch (clk_id) {
 	case CLOCK_REALTIME:
-		getnstimeofday64(&ts64);
+		ktime_get_real_ts64(&ts64);
 		tp->tv_sec = ts64.tv_sec;
 		tp->tv_nsec = ts64.tv_nsec;
 		dprintk("%s: CLOCK_REALTIME,%ld.%09ld\n", __func__,
@@ -1061,12 +1061,14 @@ static int futex(uint32_t *uaddr, int op, uint32_t val, uint64_t timeout,
 	switch (cmd) {
 	case FUTEX_WAIT:
 		val3 = FUTEX_BITSET_MATCH_ANY;
+		/* fallthrough */
 	case FUTEX_WAIT_BITSET:
 		ret = futex_wait(uaddr, fshared, val, timeout,
 				val3, clockrt, uti_info);
 		break;
 	case FUTEX_WAKE:
 		val3 = FUTEX_BITSET_MATCH_ANY;
+		/* fallthrough */
 	case FUTEX_WAKE_BITSET:
 		ret = futex_wake(uaddr, fshared, val, val3, uti_info);
 		break;
@@ -1127,7 +1129,7 @@ long do_futex(int n, unsigned long arg0, unsigned long arg1,
 	int op = (int)arg1;
 	uint32_t val = (uint32_t)arg2;
 	struct timespec *utime = (struct timespec *)arg3;
-	struct timespec ts;
+	struct timespec64 ts;
 	uint32_t *uaddr2 = (uint32_t *)arg4;
 	uint32_t val3 = (uint32_t)arg5;
 	int flags = op;
@@ -1158,12 +1160,12 @@ long do_futex(int n, unsigned long arg0, unsigned long arg1,
 		}
 
 		dprintk("%s: utime=%ld.%09ld\n", __func__, ts.tv_sec, ts.tv_nsec);
-		if (!timespec_valid(&ts)) {
+		if (!timespec64_valid(&ts)) {
 			return -EINVAL;
 		}
 
 		if (op == FUTEX_WAIT_BITSET) { /* User passed absolute time */
-			struct timespec ats;
+			struct timespec64 ats;
 
 			ret = uti_clock_gettime((flags & FUTEX_CLOCK_REALTIME) ?
 					CLOCK_REALTIME : CLOCK_MONOTONIC, &ats);

executer/kernel/mcctrl/kallsyms_compat.h

@@ -0,0 +1,72 @@
+/* kallsyms_compat.h - Compatibility layer for kallsyms_lookup_name */
+
+#ifndef KALLSYMS_COMPAT_H
+#define KALLSYMS_COMPAT_H
+
+#include <linux/version.h>
+#include <linux/kallsyms.h>
+#include <linux/kprobes.h>
+
+/* kallsyms_lookup_name is no longer exported since kernel 5.7 */
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 7, 0)
+
+/* Function pointer for kallsyms_lookup_name */
+static unsigned long (*mcctrl_kallsyms_lookup_name)(const char *name);
+
+/* Kprobe-based approach to get kallsyms_lookup_name function pointer */
+static struct kprobe kp_kallsyms = {
+	.symbol_name = "kallsyms_lookup_name"
+};
+
+static int init_kallsyms_lookup(void)
+{
+	int ret;
+	
+	ret = register_kprobe(&kp_kallsyms);
+	if (ret < 0) {
+		pr_err("register_kprobe failed, returned %d\n", ret);
+		return ret;
+	}
+	
+	mcctrl_kallsyms_lookup_name = (unsigned long (*)(const char *))kp_kallsyms.addr;
+	unregister_kprobe(&kp_kallsyms);
+	
+	if (!mcctrl_kallsyms_lookup_name) {
+		pr_err("Failed to get kallsyms_lookup_name address\n");
+		return -EINVAL;
+	}
+	
+	return 0;
+}
+
+static void cleanup_kallsyms_lookup(void)
+{
+	mcctrl_kallsyms_lookup_name = NULL;
+}
+
+static inline unsigned long mcctrl_lookup_name(const char *name)
+{
+	if (mcctrl_kallsyms_lookup_name)
+		return mcctrl_kallsyms_lookup_name(name);
+	return 0;
+}
+
+#else /* LINUX_VERSION_CODE < KERNEL_VERSION(5, 7, 0) */
+
+static inline int init_kallsyms_lookup(void)
+{
+	return 0;
+}
+
+static inline void cleanup_kallsyms_lookup(void)
+{
+}
+
+static inline unsigned long mcctrl_lookup_name(const char *name)
+{
+	return kallsyms_lookup_name(name);
+}
+
+#endif /* LINUX_VERSION_CODE >= KERNEL_VERSION(5, 7, 0) */
+
+#endif /* KALLSYMS_COMPAT_H */

executer/kernel/mcctrl/procfs.c

@@ -40,7 +40,7 @@ typedef gid_t kgid_t;
 struct procfs_entry {
 	char *name;
 	mode_t mode;
-	const struct file_operations *fops;
+	const struct proc_ops *fops;
 };
 
 #define NOD(NAME, MODE, FOP) {				\
@@ -58,8 +58,8 @@ struct procfs_entry {
 static const struct procfs_entry tid_entry_stuff[];
 static const struct procfs_entry pid_entry_stuff[];
 static const struct procfs_entry base_entry_stuff[];
-static const struct file_operations mckernel_forward_ro;
-static const struct file_operations mckernel_forward;
+static const struct proc_ops mckernel_forward_ro;
+static const struct proc_ops mckernel_forward;
 
 static ssize_t mckernel_procfs_read(struct file *file, char __user *buf, 
 		size_t nbytes, loff_t *ppos);
@@ -84,7 +84,7 @@ struct procfs_list_entry {
  * file.
  */
 LIST_HEAD(procfs_file_list);
-DEFINE_SEMAPHORE(procfs_file_list_lock);
+DEFINE_SEMAPHORE(procfs_file_list_lock, 1);
 
 static char *
 getpath(struct procfs_list_entry *e, char *buf, int bufsize)
@@ -183,10 +183,10 @@ add_procfs_entry(struct procfs_list_entry *parent, const char *name, int mode,
 		pde = proc_symlink(name, parent_pde, (char *)opaque);
 	}
 	else {
-		const struct file_operations *fop;
+		const struct proc_ops *fop;
 
 		if(opaque)
-			fop = (const struct file_operations *)opaque;
+			fop = (const struct proc_ops *)opaque;
 		else if(mode & S_IWUSR)
 			fop = &mckernel_forward;
 		else
@@ -509,7 +509,7 @@ static ssize_t __mckernel_procfs_read_write(
 	struct proc_dir_entry *dp = PDE(inode);
 	struct procfs_list_entry *e = dp->data;
 #else
-	struct procfs_list_entry *e = PDE_DATA(inode);
+	struct procfs_list_entry *e = pde_data(inode);
 #endif
 	loff_t offset = *ppos;
 	char pathbuf[PROCFS_NAME_MAX];
@@ -770,16 +770,16 @@ int procfsm_packet_handler(void *os, int msg, int pid, unsigned long arg,
 	return 0;
 }
 
-static const struct file_operations mckernel_forward_ro = {
-	.llseek		= mckernel_procfs_lseek,
-	.read		= mckernel_procfs_read,
-	.write		= NULL,
+static const struct proc_ops mckernel_forward_ro = {
+	.proc_lseek	= mckernel_procfs_lseek,
+	.proc_read	= mckernel_procfs_read,
+	.proc_write	= NULL,
 };
 
-static const struct file_operations mckernel_forward = {
-	.llseek		= mckernel_procfs_lseek,
-	.read		= mckernel_procfs_read,
-	.write		= mckernel_procfs_write,
+static const struct proc_ops mckernel_forward = {
+	.proc_lseek	= mckernel_procfs_lseek,
+	.proc_read	= mckernel_procfs_read,
+	.proc_write	= mckernel_procfs_write,
 };
 
 #define PA_NULL (-1L)
@@ -812,7 +812,7 @@ static int mckernel_procfs_buff_open(struct inode *inode, struct file *file)
 	struct proc_dir_entry *dp = PDE(inode);
 	struct procfs_list_entry *e = dp->data;
 #else
-	struct procfs_list_entry *e = PDE_DATA(inode);
+	struct procfs_list_entry *e = pde_data(inode);
 #endif
 
 	os = osnum_to_os(e->osnum);
@@ -1071,12 +1071,12 @@ rep:
 	return l;
 }
 
-static const struct file_operations mckernel_buff_io = {
-	.llseek		= mckernel_procfs_lseek,
-	.read		= mckernel_procfs_buff_read,
-	.write		= NULL,
-	.open		= mckernel_procfs_buff_open,
-	.release	= mckernel_procfs_buff_release,
+static const struct proc_ops mckernel_buff_io = {
+	.proc_lseek	= mckernel_procfs_lseek,
+	.proc_read	= mckernel_procfs_buff_read,
+	.proc_write	= NULL,
+	.proc_open	= mckernel_procfs_buff_open,
+	.proc_release	= mckernel_procfs_buff_release,
 };
 
 static const struct procfs_entry tid_entry_stuff[] = {

executer/kernel/mcctrl/syscall.c

@@ -57,6 +57,21 @@
 #include <archdeps.h>
 #include <asm/pgtable.h>
 
+/* Compatibility function for vfs_fstat which is not exported in newer kernels */
+static inline int mcctrl_vfs_fstat(int fd, struct kstat *stat)
+{
+	struct file *file;
+	int error;
+
+	file = fget(fd);
+	if (!file)
+		return -EBADF;
+
+	error = vfs_getattr(&file->f_path, stat, STATX_BASIC_STATS, AT_STATX_SYNC_AS_STAT);
+	fput(file);
+	return error;
+}
+
 #define ALIGN_WAIT_BUF(z)   (((z + 63) >> 6) << 6)
 
 //#define SC_DEBUG
@@ -733,7 +748,7 @@ static struct vm_operations_struct rus_vmops = {
 
 static int rus_mmap(struct file *file, struct vm_area_struct *vma)
 {
-	vma->vm_flags |= arch_rus_vm_flags;
+	vm_flags_set(vma, arch_rus_vm_flags);
 	vma->vm_ops = &rus_vmops;
 	return 0;
 }
@@ -776,10 +791,10 @@ reserve_user_space_common(struct mcctrl_usrdata *usrdata, unsigned long start, u
 #if 0
 	{ /* debug */
         struct vm_area_struct *vma;
-		down_write(&current->mm->mmap_sem);
+		mmap_write_lock(current->mm);
 		vma = find_vma(current->mm, start);
-		vma->vm_flags |= VM_DONTCOPY;
-		up_write(&current->mm->mmap_sem);
+		vm_flags_set(vma, VM_DONTCOPY);
+		mmap_write_unlock(current->mm);
 	}
 #endif
 	revert_creds(original);
@@ -950,7 +965,7 @@ static int pager_req_create(ihk_os_t os, int fd, uintptr_t result_pa)
 
 	dprintk("pager_req_create(%d,%lx)\n", fd, (long)result_pa);
 
-	error = vfs_fstat(fd, &st);
+	error = mcctrl_vfs_fstat(fd, &st);
 	if (error) {
 		printk("pager_req_create(%d,%lx):vfs_stat failed. %d\n", fd, (long)result_pa, error);
 		goto out;
@@ -1463,12 +1478,12 @@ static int pager_req_map(ihk_os_t os, int fd, size_t len, off_t off,
 	}
 
 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,5,0)
-	down_write(&current->mm->mmap_sem);
+	mmap_write_lock(current->mm);
 
 	va = do_mmap_pgoff(file, ANY_WHERE, len, maxprot, 
 			prot_and_flags, pgoff);
 
-	up_write(&current->mm->mmap_sem);
+	mmap_write_unlock(current->mm);
 #else
 	va = vm_mmap(file, ANY_WHERE, len, maxprot,
 			prot_and_flags, pgoff << PAGE_SHIFT);
@@ -1571,7 +1586,7 @@ static int pager_req_pfn(ihk_os_t os, uintptr_t handle, off_t off, uintptr_t ppf
 #define	PFN_VALID	((uintptr_t)1 << 63)
 	pfn = PFN_VALID;	/* Use "not present" as the default setting */
 
-	down_read(&current->mm->mmap_sem);
+	mmap_read_lock(current->mm);
 retry:	
 	pgd = pgd_offset(current->mm, va);
 	if (!pgd_none(*pgd) && !pgd_bad(*pgd) && pgd_present(*pgd)) {
@@ -1587,7 +1602,7 @@ retry:
 				pmd = pmd_offset(pud, va);
 				if (!pmd_none(*pmd) && !pmd_bad(*pmd) &&
 				    pmd_present(*pmd)) {
-					pte = pte_offset_map(pmd, va);
+					pte = pte_offset_kernel(pmd, va);
 					if (!pte_none(*pte) && pte_present(*pte)) {
 						pfn = (uintptr_t)pte_pfn(*pte) << PAGE_SHIFT;
 #define	PFN_PRESENT	((uintptr_t)1 << 0)
@@ -1623,7 +1638,9 @@ retry:
 			goto out_release;
 		}
 
-#if LINUX_VERSION_CODE >= KERNEL_VERSION(4, 8, 0) || \
+#if LINUX_VERSION_CODE >= KERNEL_VERSION(5, 1, 0)
+		fault = handle_mm_fault(vma, va, flags, NULL);
+#elif LINUX_VERSION_CODE >= KERNEL_VERSION(4, 8, 0) || \
 	(defined(RHEL_RELEASE_CODE) && RHEL_RELEASE_CODE >= RHEL_RELEASE_VERSION(7, 5))
 		fault = handle_mm_fault(vma, va, flags);
 #else
@@ -1656,7 +1673,7 @@ retry:
 	}
 
 out_release:
-	up_read(&current->mm->mmap_sem);
+	mmap_read_unlock(current->mm);
 
 	phys = ihk_device_map_memory(dev, ppfn_rpa, sizeof(*ppfn));
 	ppfn = ihk_device_map_virtual(dev, phys, sizeof(*ppfn), NULL, 0);
@@ -1682,9 +1699,9 @@ static int __pager_unmap(struct pager *pager)
 	int error;
 
 #if LINUX_VERSION_CODE < KERNEL_VERSION(3,5,0)
-	down_write(&current->mm->mmap_sem);
+	mmap_write_lock(current->mm);
 	error = do_munmap(current->mm, pager->map_uaddr, pager->map_len);
-	up_write(&current->mm->mmap_sem);
+	mmap_write_unlock(current->mm);
 #else
 	error = vm_munmap(pager->map_uaddr, pager->map_len);
 #endif
@@ -1749,9 +1766,11 @@ static long pager_req_mlock_list(ihk_os_t os, unsigned long start,
 	struct vm_area_struct	*vma;
 
 	kprintf("pager_req_mlock_list: addr(%p)\n", addr);
-	vma = find_vma(current->mm, 0x7010a0);
-	for (vma = mm->mmap; vma != NULL; vma = vma->vm_next) {
-		if (vma->vm_start < start || vma->vm_start > end) continue;
+	
+	/* Use find_vma to iterate through VMAs */
+	vma = find_vma(mm, start);
+	while (vma != NULL) {
+		if (vma->vm_start > end) break;
 		kprintf("\t%p: %p -- %p\t%lx\n", vma,
 			(void*)vma->vm_start, (void*)vma->vm_end,
 			vma->vm_flags & VM_LOCKED);
@@ -1766,6 +1785,8 @@ static long pager_req_mlock_list(ihk_os_t os, unsigned long start,
 			addrpair->flag = vma->vm_flags;
 			addrpair++;
 		}
+		/* Use find_vma to get next VMA */
+		vma = find_vma(mm, vma->vm_end);
 	}
 full:
 	return cnt;
@@ -2129,14 +2150,14 @@ static int remap_user_space(uintptr_t rva, size_t len, int prot)
 	uintptr_t map;
 
 	dprintk("remap_user_space(%lx,%lx,%x)\n", rva, len, prot);
-	down_write(&mm->mmap_sem);
+	mmap_write_lock(mm);
 	vma = find_vma(mm, rva);
 	if (!vma || (rva < vma->vm_start)) {
 		printk("remap_user_space(%lx,%lx,%x):find_vma failed. %p %lx %lx\n",
 				rva, len, prot, vma,
 				(vma)? vma->vm_start: -1,
 				(vma)? vma->vm_end: 0);
-		up_write(&mm->mmap_sem);
+		mmap_write_unlock(mm);
 		map = -ENOMEM;
 		goto out;
 	}
@@ -2150,7 +2171,7 @@ static int remap_user_space(uintptr_t rva, size_t len, int prot)
 			prot, MAP_FIXED|MAP_SHARED, pgoff);
 #endif
 
-	up_write(&mm->mmap_sem);
+	mmap_write_unlock(mm);
 
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(3,5,0)
 	map = vm_mmap(file, start, len,
@@ -2175,7 +2196,7 @@ int mcctrl_clear_pte_range(uintptr_t start, uintptr_t len)
 	int ret;
 
 	ret = 0;
-	down_read(&mm->mmap_sem);
+	mmap_read_lock(mm);
 	addr = start;
 	while (addr < (start + len)) {
 		vma = find_vma(mm, addr);
@@ -2193,7 +2214,7 @@ int mcctrl_clear_pte_range(uintptr_t start, uintptr_t len)
 		if (addr < end) {
 #if LINUX_VERSION_CODE < KERNEL_VERSION(4, 18, 0)
 			/* Revert permission */
-			vma->vm_flags |= VM_READ | VM_WRITE | VM_EXEC;
+			vm_flags_set(vma, VM_READ | VM_WRITE | VM_EXEC);
 			error = zap_vma_ptes(vma, addr, end-addr);
 			if (error) {
 				mcctrl_zap_page_range(vma, addr, end-addr,
@@ -2212,14 +2233,14 @@ int mcctrl_clear_pte_range(uintptr_t start, uintptr_t len)
 			}
 			else {
 				/* Revert permission */
-				vma->vm_flags |= VM_READ | VM_WRITE | VM_EXEC;
+				vm_flags_set(vma, VM_READ | VM_WRITE | VM_EXEC);
 				zap_vma_ptes(vma, addr, end-addr);
 			}
 #endif /* LINUX_VERSION_CODE < KERNEL_VERSION(4, 18, 0) */
 		}
 		addr = end;
 	}
-	up_read(&mm->mmap_sem);
+	mmap_read_unlock(mm);
 	return ret;
 }
 

executer/user/CMakeLists.txt

@@ -20,20 +20,26 @@ target_include_directories(mcexec PUBLIC "${KERNEL_DIR}")
 set_property(TARGET mcexec PROPERTY POSITION_INDEPENDENT_CODE ON)
 set_property(TARGET mcexec PROPERTY LINK_FLAGS "-fPIE -pie")
 
+#unset(LIBDWARF CACHE)
 add_executable(mcinspect mcinspect.c)
-if (NOT LIBDWARF)
-	target_include_directories(mcinspect PRIVATE "${CMAKE_CURRENT_BINARY_DIR}/lib/")
-	target_include_directories(mcinspect PRIVATE
-		"${CMAKE_CURRENT_SOURCE_DIR}/lib/libdwarf/libdwarf/libdwarf/")
-	target_link_libraries(mcinspect dwarf z elf)
-else()
-	target_include_directories(mcinspect PRIVATE ${DWARF_H})
-	target_link_libraries(mcinspect ${LIBDWARF})
-endif()
-target_link_libraries(mcinspect ${LIBBFD})
+target_link_libraries(mcinspect PRIVATE bfd dwarf z elf)
+#if (NOT LIBDWARF)
+#	target_include_directories(mcinspect PRIVATE "${CMAKE_CURRENT_BINARY_DIR}/lib/")
+#	target_include_directories(mcinspect PRIVATE
+#		"${CMAKE_CURRENT_SOURCE_DIR}/lib/libdwarf/libdwarf/libdwarf/")
+#	target_link_libraries(mcinspect dwarf z elf)
+#else()
+#	target_include_directories(mcinspect PRIVATE ${DWARF_H})
+#	target_link_libraries(mcinspect ${LIBDWARF})
+#endif()
+#target_link_libraries(mcinspect ${LIBBFD})
+set_property(TARGET mcinspect PROPERTY POSITION_INDEPENDENT_CODE ON)
+set_property(TARGET mcinspect PROPERTY LINK_FLAGS "-fPIE -pie")
 
 add_executable(eclair eclair.c arch/${ARCH}/arch-eclair.c)
 target_link_libraries(eclair ${LIBBFD})
+set_property(TARGET eclair PROPERTY POSITION_INDEPENDENT_CODE ON)
+set_property(TARGET eclair PROPERTY LINK_FLAGS "-fPIE -pie")
 
 add_library(sched_yield SHARED libsched_yield.c)
 target_link_libraries(sched_yield dl)

executer/user/ldump2mcdump.c

@@ -279,13 +279,13 @@ void cmd_ldump2mcdump(void)
 			return;
 		}
 
-		ok = bfd_set_section_size(abfd, scn, cpsize);
+		ok = bfd_set_section_size(scn, cpsize);
 		if (!ok) {
 			bfd_perror("bfd_set_section_size");
 			return;
 		}
 
-		ok = bfd_set_section_flags(abfd, scn, SEC_HAS_CONTENTS);
+		ok = bfd_set_section_flags(scn, SEC_HAS_CONTENTS);
 		if (!ok) {
 			bfd_perror("bfd_set_setction_flags");
 			return;
@@ -300,13 +300,13 @@ void cmd_ldump2mcdump(void)
 			return;
 		}
 
-		ok = bfd_set_section_size(abfd, scn, cpsize);
+		ok = bfd_set_section_size(scn, cpsize);
 		if (!ok) {
 			bfd_perror("bfd_set_section_size");
 			return;
 		}
 
-		ok = bfd_set_section_flags(abfd, scn, SEC_HAS_CONTENTS);
+		ok = bfd_set_section_flags(scn, SEC_HAS_CONTENTS);
 		if (!ok) {
 			bfd_perror("bfd_set_setction_flags");
 			return;
@@ -321,13 +321,13 @@ void cmd_ldump2mcdump(void)
 			return;
 		}
 
-		ok = bfd_set_section_size(abfd, scn, cpsize);
+		ok = bfd_set_section_size(scn, cpsize);
 		if (!ok) {
 			bfd_perror("bfd_set_section_size");
 			return;
 		}
 
-		ok = bfd_set_section_flags(abfd, scn, SEC_HAS_CONTENTS);
+		ok = bfd_set_section_flags(scn, SEC_HAS_CONTENTS);
 		if (!ok) {
 			bfd_perror("bfd_set_setction_flags");
 			return;
@@ -341,13 +341,13 @@ void cmd_ldump2mcdump(void)
 		return;
 	}
 
-	ok = bfd_set_section_size(abfd, scn, mem_size);
+	ok = bfd_set_section_size(scn, mem_size);
 	if (!ok) {
 		bfd_perror("bfd_set_section_size");
 		return;
 	}
 
-	ok = bfd_set_section_flags(abfd, scn, SEC_ALLOC|SEC_HAS_CONTENTS);
+	ok = bfd_set_section_flags(scn, SEC_ALLOC|SEC_HAS_CONTENTS);
 	if (!ok) {
 		bfd_perror("bfd_set_setction_flags");
 		return;
@@ -366,14 +366,14 @@ void cmd_ldump2mcdump(void)
 			return;
 		}
 
-		ok = bfd_set_section_size(abfd, scn, mem_chunks->chunks[i].size);
+		ok = bfd_set_section_size(scn, mem_chunks->chunks[i].size);
 
 		if (!ok) {
 			bfd_perror("bfd_set_section_size");
 			return;
 		}
 
-		ok = bfd_set_section_flags(abfd, scn, SEC_ALLOC|SEC_HAS_CONTENTS);
+		ok = bfd_set_section_flags(scn, SEC_ALLOC|SEC_HAS_CONTENTS);
 		if (!ok) {
 			bfd_perror("bfd_set_setction_flags");
 			return;

executer/user/lib/CMakeLists.txt

@@ -4,9 +4,9 @@ endif()
 
 if (ENABLE_UTI)
 	if (${ARCH} STREQUAL "arm64")
-	  set(SYSCALL_INTERCEPT_SOURCE_DIR "${CMAKE_CURRENT_SOURCE_DIR}/syscall_intercept/arch/aarch64" CACHE STRINGS "relative path to syscalL_intercept source directory")
+	  set(SYSCALL_INTERCEPT_SOURCE_DIR "${CMAKE_CURRENT_SOURCE_DIR}/syscall_intercept/arch/aarch64" CACHE STRING "relative path to syscalL_intercept source directory")
 	elseif (${ARCH} STREQUAL "x86_64")
-	  set(SYSCALL_INTERCEPT_SOURCE_DIR "${CMAKE_CURRENT_SOURCE_DIR}/syscall_intercept" CACHE STRINGS "relative path to syscalL_intercept source directory")
+	  set(SYSCALL_INTERCEPT_SOURCE_DIR "${CMAKE_CURRENT_SOURCE_DIR}/syscall_intercept" CACHE STRING "relative path to syscalL_intercept source directory")
 	endif()
 
 

executer/user/mcinspect.c

@@ -19,7 +19,7 @@
 #include <string.h>
 #include <errno.h>
 #include <dwarf.h>
-#include <libdwarf/libdwarf.h>
+#include <libdwarf-0/libdwarf.h>
 #include <getopt.h>
 #include <libgen.h>
 #include <bfd.h>
@@ -130,14 +130,15 @@ int dwarf_walk_tree(Dwarf_Debug dbg,
 	Dwarf_Error err;
 	Dwarf_Die unit;
 	Dwarf_Die die;
+	Dwarf_Half header_cu_type;
 	int rc;
 
 	/* Iterate compile and type units */
 	for (is_info = 0; is_info < 2; ++is_info) {
-		rc = dwarf_next_cu_header_c(dbg, is_info, &cu_length,
+		rc = dwarf_next_cu_header_d(dbg, is_info, &cu_length,
 				&cu_version, &cu_abbrev_offset, &cu_pointer_size,
 				&cu_offset_size, &cu_extension_size, &type_signature,
-				&type_offset, &cu_next_offset, &err);
+				&type_offset, &cu_next_offset, &header_cu_type, &err);
 
 		while (rc != DW_DLV_NO_ENTRY) {
 			char *name = NULL;
@@ -151,9 +152,9 @@ int dwarf_walk_tree(Dwarf_Debug dbg,
 				return -1;
 			}
 
-			rc = dwarf_siblingof(dbg, NULL, &unit, &err);
+			rc = dwarf_siblingof_b(dbg, NULL, is_info, &unit, &err);
 			if (rc != DW_DLV_OK) {
-				fprintf(stderr, "error: dwarf_siblingof failed: %d %s\n",
+				fprintf(stderr, "error: dwarf_siblingof_b failed: %d %s\n",
 						rc, dwarf_errmsg(err));
 				return -1;
 			}
@@ -237,7 +238,7 @@ int dwarf_walk_tree(Dwarf_Debug dbg,
 					}
 				}
 
-				rc = dwarf_siblingof(dbg, die, &next, &err);
+				rc = dwarf_siblingof_b(dbg, die, is_info, &next, &err);
 				dwarf_dealloc(dbg, die, DW_DLA_DIE);
 				if (name)
 					dwarf_dealloc(dbg, name, DW_DLA_STRING);
@@ -248,10 +249,10 @@ int dwarf_walk_tree(Dwarf_Debug dbg,
 				die = next;
 			}
 
-			rc = dwarf_next_cu_header_c(dbg, is_info, &cu_length,
+			rc = dwarf_next_cu_header_d(dbg, is_info, &cu_length,
 					&cu_version, &cu_abbrev_offset, &cu_pointer_size,
 					&cu_offset_size, &cu_extension_size, &type_signature,
-					&type_offset, &cu_next_offset, &err);
+					&type_offset, &cu_next_offset, &header_cu_type, &err);
 		}
 	}
 
@@ -294,34 +295,64 @@ int dwarf_get_size(Dwarf_Debug dbg,
 
 		if (ssize < 0) {
 			fprintf(stderr, "%s: unsupported negative size\n",
-				__func__);
+					__func__);
 			return DW_DLV_ERROR;
 		}
 
 		size = (Dwarf_Unsigned) ssize;
 	}
 	else {
-		Dwarf_Locdesc **locdescs;
-		Dwarf_Signed len;
+		Dwarf_Loc_Head_c loclist_head = 0;
+		Dwarf_Unsigned lcount = 0;
+		Dwarf_Locdesc_c locdesc_entry = 0;
+		Dwarf_Small op;
+		Dwarf_Unsigned opd1, opd2, opd3;
+		Dwarf_Unsigned offsetforbranch;
+		int lres;
 
-		if (dwarf_loclist_n(attr, &locdescs, &len,  perr)
-				== DW_DLV_ERROR) {
+
+		lres = dwarf_get_loclist_c(attr, &loclist_head, &lcount, perr);
+		if (lres != DW_DLV_OK) {
 			fprintf(stderr, "%s: unsupported member size\n",
 					__func__);
 			return DW_DLV_ERROR;
 		}
 
-		if (len != 1 ||
-				locdescs[0]->ld_cents != 1 ||
-				(locdescs[0]->ld_s[0]).lr_atom
-				!= DW_OP_plus_uconst) {
+		if (lcount != 1) {
 			fprintf(stderr,
 					"%s: unsupported location expression\n",
 					__func__);
+			dwarf_loc_head_c_dealloc(loclist_head);
 			return DW_DLV_ERROR;
 		}
 
-		size = (locdescs[0]->ld_s[0]).lr_number;
+		lres = dwarf_get_locdesc_entry_d(loclist_head, 0, 0, 0, 0, 0, 0, 0, 0, &locdesc_entry, 0, 0, 0, perr);
+		if (lres != DW_DLV_OK) {
+			fprintf(stderr,
+					"%s: unsupported location expression\n",
+					__func__);
+			dwarf_loc_head_c_dealloc(loclist_head);
+			return DW_DLV_ERROR;
+		}
+
+		lres = dwarf_get_location_op_value_d(locdesc_entry, 0, &op, &opd1, &opd2, &opd3, NULL, NULL, NULL, &offsetforbranch, perr);
+		if (lres != DW_DLV_OK) {
+			fprintf(stderr,
+					"%s: unsupported location expression\n",
+					__func__);
+			dwarf_loc_head_c_dealloc(loclist_head);
+			return DW_DLV_ERROR;
+		}
+
+		if (op != DW_OP_plus_uconst) {
+			fprintf(stderr,
+					"%s: unsupported location expression\n",
+					__func__);
+			dwarf_loc_head_c_dealloc(loclist_head);
+			return DW_DLV_ERROR;
+		}
+		size = opd1;
+		dwarf_loc_head_c_dealloc(loclist_head);
 	}
 	dwarf_dealloc(dbg, attr, DW_DLA_ATTR);
 
@@ -455,27 +486,57 @@ int dwarf_get_offset(Dwarf_Debug dbg,
 		offset = (Dwarf_Unsigned) soffset;
 	}
 	else {
-		Dwarf_Locdesc **locdescs;
-		Dwarf_Signed len;
+		Dwarf_Loc_Head_c loclist_head = 0;
+		Dwarf_Unsigned lcount = 0;
+		Dwarf_Locdesc_c locdesc_entry = 0;
+		Dwarf_Small op;
+		Dwarf_Unsigned opd1, opd2, opd3;
+		Dwarf_Unsigned offsetforbranch;
+		int lres;
 
-		if (dwarf_loclist_n(attr, &locdescs, &len,  perr)
-				== DW_DLV_ERROR) {
+
+		lres = dwarf_get_loclist_c(attr, &loclist_head, &lcount, perr);
+	if (lres != DW_DLV_OK) {
 			fprintf(stderr, "%s: unsupported member offset\n",
 					__func__);
 			return DW_DLV_ERROR;
 		}
 
-		if (len != 1 ||
-				locdescs[0]->ld_cents != 1 ||
-				(locdescs[0]->ld_s[0]).lr_atom
-				!= DW_OP_plus_uconst) {
+		if (lcount != 1) {
 			fprintf(stderr,
 					"%s: unsupported location expression\n",
 					__func__);
+			dwarf_loc_head_c_dealloc(loclist_head);
 			return DW_DLV_ERROR;
 		}
 
-		offset = (locdescs[0]->ld_s[0]).lr_number;
+		lres = dwarf_get_locdesc_entry_d(loclist_head, 0, 0, 0, 0, 0, 0, 0, 0, &locdesc_entry, 0, 0, 0, perr);
+		if (lres != DW_DLV_OK) {
+			fprintf(stderr,
+					"%s: unsupported location expression\n",
+					__func__);
+			dwarf_loc_head_c_dealloc(loclist_head);
+			return DW_DLV_ERROR;
+		}
+
+		lres = dwarf_get_location_op_value_d(locdesc_entry, 0, &op, &opd1, &opd2, &opd3, NULL, NULL, NULL, &offsetforbranch, perr);
+		if (lres != DW_DLV_OK) {
+			fprintf(stderr,
+					"%s: unsupported location expression\n",
+					__func__);
+			dwarf_loc_head_c_dealloc(loclist_head);
+			return DW_DLV_ERROR;
+		}
+
+		if (op != DW_OP_plus_uconst) {
+			fprintf(stderr,
+					"%s: unsupported location expression\n",
+					__func__);
+			dwarf_loc_head_c_dealloc(loclist_head);
+			return DW_DLV_ERROR;
+		}
+		offset = opd1;
+		dwarf_loc_head_c_dealloc(loclist_head);
 	}
 	dwarf_dealloc(dbg, attr, DW_DLA_ATTR);
 
@@ -579,10 +640,10 @@ int dwarf_struct_field_offset(Dwarf_Debug dbg, Dwarf_Die die, void *arg)
 		break;
 
 next_child:
-		rc = dwarf_siblingof(dbg, child, &next, &err);
+		rc = dwarf_siblingof_b(dbg, child, 1, &next, &err);
 		dwarf_dealloc(dbg, child, DW_DLA_DIE);
 		if (rc != DW_DLV_OK) {
-			fprintf(stderr, "%s: error: dwarf_siblingof: %d %s\n",
+			fprintf(stderr, "%s: error: dwarf_siblingof_b: %d %s\n",
 					__func__, rc, dwarf_errmsg(err));
 			rc = DW_DLV_NO_ENTRY;
 			goto out;
@@ -617,7 +678,7 @@ out:
 	rc = dwarf_walk_tree(dbg, dwarf_struct_field_offset, &dsfo); \
 	if (rc != DW_DLV_OK) { \
 		fprintf(stderr, "%s: error: finding %s in struct %s\n", \
-		   __func__, dsfo.field_name, dsfo.struct_name); \
+			__func__, dsfo.field_name, dsfo.struct_name); \
 	   exit(1); \
 	} \
 	offset; \
@@ -681,7 +742,7 @@ int dwarf_get_address(Dwarf_Debug dbg,
 		printf("%s: DW_AT_location\n", __func__);
 
 		rc = dwarf_whatform(attr, &form, perr);
-		if (rc != DW_DLV_OK) {
+	if (rc != DW_DLV_OK) {
 			fprintf(stderr, "%s: error: getting whatform: %s\n",
 					__func__, dwarf_errmsg(*perr));
 			goto dealloc_out;
@@ -696,90 +757,148 @@ int dwarf_get_address(Dwarf_Debug dbg,
 				form == DW_FORM_data8 ||
 				form == DW_FORM_sec_offset) {
 
-			Dwarf_Locdesc **locdescs;
-			Dwarf_Signed len;
+			Dwarf_Loc_Head_c loclist_head = 0;
+			Dwarf_Unsigned lcount = 0;
+			Dwarf_Locdesc_c locdesc_entry = 0;
+			Dwarf_Small op;
+			Dwarf_Unsigned opd1, opd2, opd3;
+			Dwarf_Unsigned offsetforbranch;
+			int lres;
 
-			if (dwarf_loclist_n(attr, &locdescs, &len,  perr)
-					== DW_DLV_ERROR) {
-				fprintf(stderr, "%s: dwarf_loclist_n: %s\n",
+			lres = dwarf_get_loclist_c(attr, &loclist_head, &lcount, perr);
+			if (lres != DW_DLV_OK) {
+				fprintf(stderr, "%s: dwarf_get_loclist_c: %s\n",
 						__func__, dwarf_errmsg(*perr));
 				rc = DW_DLV_ERROR;
 				goto dealloc_out;
 			}
 
-			if (len != 1 ||
-					locdescs[0]->ld_cents != 1 ||
-					(locdescs[0]->ld_s[0]).lr_atom
-					!= DW_OP_addr) {
+			if (lcount != 1) {
 				fprintf(stderr,
 						"%s: unsupported addr expression\n",
 						__func__);
+				dwarf_loc_head_c_dealloc(loclist_head);
 				rc = DW_DLV_ERROR;
 				goto dealloc_out;
 			}
 
-			addr = (locdescs[0]->ld_s[0]).lr_number;
+			lres = dwarf_get_locdesc_entry_d(loclist_head, 0, 0, 0, 0, 0, 0, 0, 0, &locdesc_entry, 0, 0, 0, perr);
+			if (lres != DW_DLV_OK) {
+				fprintf(stderr,
+						"%s: unsupported addr expression\n",
+						__func__);
+				dwarf_loc_head_c_dealloc(loclist_head);
+				rc = DW_DLV_ERROR;
+				goto dealloc_out;
+			}
+
+			lres = dwarf_get_location_op_value_d(locdesc_entry, 0, &op, &opd1, &opd2, &opd3, NULL, NULL, NULL, &offsetforbranch, perr);
+			if (lres != DW_DLV_OK) {
+				fprintf(stderr,
+						"%s: unsupported addr expression\n",
+						__func__);
+				dwarf_loc_head_c_dealloc(loclist_head);
+				rc = DW_DLV_ERROR;
+				goto dealloc_out;
+			}
+
+			if (op != DW_OP_addr) {
+				fprintf(stderr,
+						"%s: unsupported addr expression\n",
+						__func__);
+				dwarf_loc_head_c_dealloc(loclist_head);
+				rc = DW_DLV_ERROR;
+				goto dealloc_out;
+			}
+			addr = opd1;
+			dwarf_loc_head_c_dealloc(loclist_head);
 		}
 		else if (form == DW_FORM_exprloc)  {
 			Dwarf_Half address_size = 0;
 			Dwarf_Ptr x = 0;
 			Dwarf_Unsigned tempud = 0;
-			Dwarf_Locdesc *locdescs = 0;
-			Dwarf_Signed len = 0;
+			Dwarf_Loc_Head_c loclist_head = 0;
+			Dwarf_Unsigned lcount = 0;
+			Dwarf_Locdesc_c locdesc_entry = 0;
+			Dwarf_Small op;
+			Dwarf_Unsigned opd1, opd2, opd3;
+			Dwarf_Unsigned offsetforbranch;
+			int lres;
+			Dwarf_Half version;
+			Dwarf_Half offset_size;
 
 			rc = dwarf_formexprloc(attr, &tempud, &x, perr);
-			if (rc == DW_DLV_NO_ENTRY) {
-				fprintf(stderr, "%s: dwarf_formexprloc: no entry?\n",
-						__func__);
-				goto dealloc_out;
-			}
-			else if (rc == DW_DLV_ERROR) {
+			if (rc != DW_DLV_OK) {
 				fprintf(stderr, "%s: dwarf_formexprloc(): %s\n",
 						__func__, dwarf_errmsg(*perr));
 				goto dealloc_out;
 			}
 
 			rc = dwarf_get_die_address_size(die, &address_size, perr);
-			if (rc == DW_DLV_NO_ENTRY) {
-				fprintf(stderr, "%s: dwarf_get_die_address_size: no entry?\n",
-						__func__);
-				goto dealloc_out;
-			}
-			else if (rc == DW_DLV_ERROR) {
+			if (rc != DW_DLV_OK) {
 				fprintf(stderr, "%s: dwarf_get_die_address_size: %s\n",
 						__func__, dwarf_errmsg(*perr));
 				goto dealloc_out;
 			}
 
-			rc = dwarf_loclist_from_expr_a(dbg, x, tempud, address_size,
-					&locdescs, &len, perr);
-			if (rc == DW_DLV_ERROR) {
-				fprintf(stderr, "%s: dwarf_loclist_from_expr_a: %s\n",
+			rc = dwarf_get_version_of_die(die, &version, &offset_size);
+			if (rc != DW_DLV_OK) {
+				fprintf(stderr, "%s: dwarf_get_version_of_die: %s\n",
 						__func__, dwarf_errmsg(*perr));
 				goto dealloc_out;
 			}
-			else if (rc == DW_DLV_NO_ENTRY) {
-				fprintf(stderr, "%s: dwarf_loclist_from_expr_a: no entry?\n",
-						__func__);
+
+			rc = dwarf_loclist_from_expr_c(dbg, x, tempud, address_size, offset_size, version,
+					&loclist_head, &lcount, perr);
+			if (rc != DW_DLV_OK) {
+				fprintf(stderr, "%s: dwarf_loclist_from_expr_c: %s\n",
+						__func__, dwarf_errmsg(*perr));
 				goto dealloc_out;
 			}
 
-			/* len is always 1 */
-			if (len != 1 ||
-					locdescs[0].ld_cents != 1 ||
-					(locdescs[0].ld_s[0]).lr_atom
-					!= DW_OP_addr) {
+			if (lcount != 1) {
 				fprintf(stderr,
 						"%s: unsupported addr expression\n",
 						__func__);
+				dwarf_loc_head_c_dealloc(loclist_head);
+				rc = DW_DLV_ERROR;
 				goto dealloc_out;
 			}
 
-			addr = (locdescs[0].ld_s[0]).lr_number;
+			lres = dwarf_get_locdesc_entry_d(loclist_head, 0, 0, 0, 0, 0, 0, 0, 0, &locdesc_entry, 0, 0, 0, perr);
+			if (lres != DW_DLV_OK) {
+				fprintf(stderr,
+						"%s: unsupported addr expression\n",
+						__func__);
+				dwarf_loc_head_c_dealloc(loclist_head);
+				rc = DW_DLV_ERROR;
+				goto dealloc_out;
+			}
+
+			lres = dwarf_get_location_op_value_d(locdesc_entry, 0, &op, &opd1, &opd2, &opd3, NULL, NULL, NULL, &offsetforbranch, perr);
+			if (lres != DW_DLV_OK) {
+				fprintf(stderr,
+						"%s: unsupported addr expression\n",
+						__func__);
+				dwarf_loc_head_c_dealloc(loclist_head);
+				rc = DW_DLV_ERROR;
+				goto dealloc_out;
+			}
+
+			if (op != DW_OP_addr) {
+				fprintf(stderr,
+						"%s: unsupported addr expression\n",
+						__func__);
+				dwarf_loc_head_c_dealloc(loclist_head);
+				rc = DW_DLV_ERROR;
+				goto dealloc_out;
+			}
+			addr = opd1;
+			dwarf_loc_head_c_dealloc(loclist_head);
 		}
 		else {
 			fprintf(stderr, "%s: unsupported form type?\n",
-					__func__);
+						__func__);
 			goto dealloc_out;
 		}
 
@@ -888,7 +1007,7 @@ out:
 		rc = dwarf_walk_tree(dbg, dwarf_global_var_addr, &gva); \
 		if (rc != DW_DLV_OK) { \
 			fprintf(stderr, "%s: error: finding addr of %s\n", \
-				__func__, gva.variable); \
+					__func__, gva.variable); \
 			exit(1); \
 		} \
 	} \
@@ -1210,25 +1329,25 @@ struct option mcinspect_options[] = {
 	{
 		.name =		"ps",
 		.has_arg =	no_argument,
-		.flag =		&ps,
+		.flag = 	&ps,
 		.val =		1,
 	},
 	{
 		.name =		"help",
 		.has_arg =	no_argument,
-		.flag =		&help,
+		.flag = 	&help,
 		.val =		1,
 	},
 	{
 		.name =		"debug",
 		.has_arg =	no_argument,
-		.flag =		&debug,
+		.flag = 	&debug,
 		.val =		1,
 	},
 	{
 		.name =		"vtop",
 		.has_arg =	no_argument,
-		.flag =		&vtop,
+		.flag = 	&vtop,
 		.val =		1,
 	},
 	{
@@ -1277,7 +1396,7 @@ int main(int argc, char **argv)
 		case 'v':
 			vtop_addr = strtoul(optarg, 0, 16);
 			if (vtop_addr == 0 ||
-					errno == EINVAL || errno == ERANGE) {
+						errno == EINVAL || errno == ERANGE) {
 				fprintf(stderr, "error: invalid VA? (expected format: 0xXXXX)\n\n");
 				usage(argv);
 				exit(1);
@@ -1324,7 +1443,7 @@ int main(int argc, char **argv)
 		exit(1);
 	}
 
-	rc = dwarf_init(dwarffd, DW_DLC_READ, errhand, errarg, &dbg, &error);
+	rc = dwarf_init_b(dwarffd, DW_DLA_WEAK, errhand, errarg, &dbg, &error);
 	if (rc != DW_DLV_OK) {
 		fprintf(stderr, "error: accessing DWARF information\n");
 		exit(1);
@@ -1339,7 +1458,7 @@ int main(int argc, char **argv)
 		mcvtop(dbg, pid, vtop_addr);
 	}
 
-	dwarf_finish(dbg, &error);
+	dwarf_finish(dbg);
 	close(dwarffd);
 	close(mcfd);
 	return 0;

kernel/devobj.c

@@ -136,6 +136,27 @@ int devobj_create(int fd, size_t len, off_t off, struct memobj **objp, int *maxp
 	error = 0;
 	*objp = to_memobj(obj);
 	*maxprotp = result.maxprot;
+
+#ifdef ENABLE_FUGAKU_HACKS
+	/* Pre-populate device file PFNs for PMIx shared mem */
+	if (!strncmp(obj->memobj.path,
+				"/var/opt/FJSVtcs/ple/daemonif", 29)) {
+		off_t offset;
+		uintptr_t phys;
+		unsigned long flag;
+
+		for (offset = 0; offset < obj->memobj.size; offset += PAGE_SIZE) {
+			if (devobj_get_page(&obj->memobj, offset, PAGE_P2ALIGN,
+						&phys, &flag, 0) < 0) {
+				kprintf("%s: WARNING: failed to populate offset %lu in %s\n",
+						__func__, offset, obj->memobj.path);
+			}
+		}
+		dkprintf("%s: pre-populated PFNs for %s, len: %lu\n",
+			__func__, obj->memobj.path, obj->memobj.size);
+	}
+#endif
+
 	obj = NULL;
 
 out:
@@ -200,6 +221,10 @@ static int devobj_get_page(struct memobj *memobj, off_t off, int p2align, uintpt
 	uintptr_t attr;
 	ihk_mc_user_context_t ctx;
 	int ix;
+	unsigned long irqstate;
+#ifdef ENABLE_FUGAKU_HACKS
+	int page_fault_attempts = 5;
+#endif
 
 	dkprintf("devobj_get_page(%p %lx,%lx,%d)\n", memobj, obj->handle, off, p2align);
 
@@ -214,8 +239,15 @@ static int devobj_get_page(struct memobj *memobj, off_t off, int p2align, uintpt
 #ifdef PROFILE_ENABLE
 	profile_event_add(PROFILE_page_fault_dev_file, PAGE_SIZE);
 #endif // PROFILE_ENABLE
+
+	irqstate = ihk_mc_spinlock_lock(&obj->pfn_table_lock);
 	pfn = obj->pfn_table[ix];
+	ihk_mc_spinlock_unlock(&obj->pfn_table_lock, irqstate);
+
 	if (!(pfn & PFN_VALID)) {
+#ifdef ENABLE_FUGAKU_HACKS
+pf_retry:
+#endif
 		ihk_mc_syscall_arg0(&ctx) = PAGER_REQ_PFN;
 		ihk_mc_syscall_arg1(&ctx) = obj->handle;
 		ihk_mc_syscall_arg2(&ctx) = off & ~(PAGE_SIZE - 1);
@@ -241,8 +273,24 @@ static int devobj_get_page(struct memobj *memobj, off_t off, int p2align, uintpt
 			pfn |= attr;
 			dkprintf("devobj_get_page(%p %lx,%lx,%d):PFN_PRESENT after %#lx\n", memobj, obj->handle, off, p2align, pfn);
 		}
+#ifdef ENABLE_FUGAKU_HACKS
+		else if (page_fault_attempts > 0) {
+			kprintf("%s(): va: 0x%lx !PFN_PRESENT for offset %lu in %s, "
+					"page_fault_attempts: %d\n",
+					__func__, virt_addr, off,
+					memobj->path ? memobj->path : "<unknown>",
+					page_fault_attempts);
+			--page_fault_attempts;
+			goto pf_retry;
+		}
+#endif
 
-		obj->pfn_table[ix] = pfn;
+		/* Update atomically if unset */
+		irqstate = ihk_mc_spinlock_lock(&obj->pfn_table_lock);
+		if (obj->pfn_table[ix] == 0) {
+			obj->pfn_table[ix] = pfn;
+		}
+		ihk_mc_spinlock_unlock(&obj->pfn_table_lock, irqstate);
 		// Don't call memory_stat_rss_add() because devobj related pages don't reside in main memory
 	}
 

kernel/profile.c

@@ -100,8 +100,13 @@ void profile_event_add(enum profile_event_type type, uint64_t tsc)
 		return;
 
 	if (!cpu_local_var(current)->profile_events) {
-		if (profile_alloc_events(cpu_local_var(current)) < 0)
+		if (type == PROFILE_mpol_alloc_missed) {
 			return;
+		}
+
+		if (profile_alloc_events(cpu_local_var(current)) < 0) {
+			return;
+		}
 	}
 
 	if (type < PROFILE_EVENT_MAX) {

kernel/syscall.c

@@ -2219,6 +2219,8 @@ straight_out:
 		}
 #endif // PROFILE_ENABLE
 		if (error == -ESRCH) {
+			int populate_flags = 0;
+
 			dkprintf("do_mmap:hit non VREG\n");
 			/*
 			 * XXX: temporary:
@@ -2230,8 +2232,21 @@ straight_out:
 				vrflags &= ~VR_MEMTYPE_MASK;
 				vrflags |= VR_MEMTYPE_UC;
 			}
+
+#ifdef ENABLE_FUGAKU_HACKS
+#ifdef ENABLE_TOFU
+			if (!strncmp("/var/opt/FJSVtcs/ple/daemonif/",
+						thread->proc->fd_path[fd], 30)) {
+				dkprintf("%s: MAP_POPULATE | MAP_LOCKED for %s\n",
+					__func__, thread->proc->fd_path[fd]);
+				populate_flags = (MAP_POPULATE | MAP_LOCKED);
+			}
+#endif
+#endif
+
 			error = devobj_create(fd, len, off, &memobj, &maxprot, 
-					prot, (flags & (MAP_POPULATE | MAP_LOCKED)));
+					prot,
+					populate_flags | (flags & (MAP_POPULATE | MAP_LOCKED)));
 
 			if (!error) {
 #ifdef PROFILE_ENABLE

scripts/mckernel.spec.in

@@ -71,6 +71,8 @@ This package contains headers and libraries required for build apps using IHK/Mc
 
 # We need to remove ld flags like relro for the final mckernel.img link, as well as remove cflags for mckernel
 # ideally mckernel should use different environment variables for the user tools and the kernel tools altogether...
+%undefine _hardened_build
+%define build_ldflags ""
 %define __global_ldflags ""
 %define optflags -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions --param=ssp-buffer-size=4 -grecord-gcc-switches -mtune=generic
 

scripts/mcreboot-smp.sh.in

@@ -119,7 +119,7 @@ if [ "${pid}" != "" ]; then
     ${SUDO} kill -9 ${pid} > /dev/null 2> /dev/null
 fi
 if [ "${redirect_kmsg}" != "0" -o "${mon_interval}" != "-1" ]; then
-    ${SBINDIR}/ihkmond -f ${facility} -k ${redirect_kmsg} -i ${mon_interval}
+    ${SUDO} ${SBINDIR}/ihkmond -f ${facility} -k ${redirect_kmsg} -i ${mon_interval}
 fi
 
 disable_irqbalance_mck() {