[RFC PATCH 28/47] lkl tools: add lklfuse

[Hajime Tazaki <thehajime@gmail.com>]

From: Octavian Purdila <tavi.purdila@gmail.com>

Add a simple fuse based program that can mount filesystem in userspace
using LKL.

Signed-off-by: Conrad Meyer <cem@FreeBSD.org>
Signed-off-by: Hajime Tazaki <thehajime@gmail.com>
Signed-off-by: Quentin Anglade <quentin.anglade@objectif-libre.com>
Signed-off-by: Rafael Gieschke <rafael.gieschke@rz.uni-freiburg.de>
Signed-off-by: Octavian Purdila <tavi.purdila@gmail.com>
---
 tools/lkl/lklfuse.c        | 658 +++++++++++++++++++++++++++++++++++++
 tools/lkl/tests/lklfuse.sh | 110 +++++++
 2 files changed, 768 insertions(+)
 create mode 100644 tools/lkl/lklfuse.c
 create mode 100755 tools/lkl/tests/lklfuse.sh

diff --git a/tools/lkl/lklfuse.c b/tools/lkl/lklfuse.c
new file mode 100644
index 000000000000..4e6c8fe250d0
--- /dev/null
+++ b/tools/lkl/lklfuse.c
@@ -0,0 +1,658 @@
+// SPDX-License-Identifier: GPL-2.0
+#include <stdbool.h>
+#include <stddef.h>
+#include <stdio.h>
+#include <string.h>
+#include <stdlib.h>
+#include <sys/stat.h>
+#include <string.h>
+#include <errno.h>
+#include <unistd.h>
+#define FUSE_USE_VERSION 26
+#include <fuse.h>
+#include <fuse/fuse_opt.h>
+#include <fuse/fuse_lowlevel.h>
+#include <lkl.h>
+#include <lkl_host.h>
+
+#define LKLFUSE_VERSION "0.1"
+
+struct lklfuse {
+	const char *file;
+	const char *log;
+	const char *type;
+	const char *opts;
+	struct lkl_disk disk;
+	int disk_id;
+	int part;
+	int ro;
+	int mb;
+} lklfuse = {
+	.mb = 64,
+};
+
+#define LKLFUSE_OPT(t, p, v) { t, offsetof(struct lklfuse, p), v }
+
+enum {
+	KEY_HELP,
+	KEY_VERSION,
+};
+
+static struct fuse_opt lklfuse_opts[] = {
+	LKLFUSE_OPT("log=%s", log, 0),
+	LKLFUSE_OPT("type=%s", type, 0),
+	LKLFUSE_OPT("mb=%d", mb, 0),
+	LKLFUSE_OPT("opts=%s", opts, 0),
+	LKLFUSE_OPT("part=%d", part, 0),
+	FUSE_OPT_KEY("-h", KEY_HELP),
+	FUSE_OPT_KEY("--help", KEY_HELP),
+	FUSE_OPT_KEY("-V",             KEY_VERSION),
+	FUSE_OPT_KEY("--version",      KEY_VERSION),
+	FUSE_OPT_END
+};
+
+static void usage(void)
+{
+	printf(
+"usage: lklfuse file mountpoint [options]\n"
+"\n"
+"general options:\n"
+"    -o opt,[opt...]        mount options\n"
+"    -h   --help            print help\n"
+"    -V   --version         print version\n"
+"\n"
+"lklfuse options:\n"
+"    -o log=FILE            log file\n"
+"    -o type=fstype         filesystem type\n"
+"    -o mb=memory in mb     ammount of memory to allocate\n"
+"    -o part=parition       partition to mount\n"
+"    -o ro                  open file read-only\n"
+"    -o opts=options        mount options (use \\ to escape , and =)\n"
+);
+}
+
+static int lklfuse_opt_proc(void *data, const char *arg, int key,
+			  struct fuse_args *args)
+{
+	switch (key) {
+	case FUSE_OPT_KEY_OPT:
+		if (strcmp(arg, "ro") == 0)
+			lklfuse.ro = 1;
+		return 1;
+
+	case FUSE_OPT_KEY_NONOPT:
+		if (!lklfuse.file) {
+			lklfuse.file = strdup(arg);
+			return 0;
+		}
+		return 1;
+
+	case KEY_HELP:
+		usage();
+		fuse_opt_add_arg(args, "-ho");
+		fuse_main(args->argc, args->argv, NULL, NULL);
+		exit(1);
+
+	case KEY_VERSION:
+		printf("lklfuse version %s\n", LKLFUSE_VERSION);
+		fuse_opt_add_arg(args, "--version");
+		fuse_main(args->argc, args->argv, NULL, NULL);
+		exit(0);
+
+	default:
+		fprintf(stderr, "internal error\n");
+		abort();
+	}
+}
+
+static void lklfuse_xlat_stat(const struct lkl_stat *in, struct stat *st)
+{
+	st->st_dev = in->st_dev;
+	st->st_ino = in->st_ino;
+	st->st_mode = in->st_mode;
+	st->st_nlink = in->st_nlink;
+	st->st_uid = in->st_uid;
+	st->st_gid = in->st_gid;
+	st->st_rdev = in->st_rdev;
+	st->st_size = in->st_size;
+	st->st_blksize = in->st_blksize;
+	st->st_blocks = in->st_blocks;
+	st->st_atim.tv_sec = in->lkl_st_atime;
+	st->st_atim.tv_nsec = in->st_atime_nsec;
+	st->st_mtim.tv_sec = in->lkl_st_mtime;
+	st->st_mtim.tv_nsec = in->st_mtime_nsec;
+	st->st_ctim.tv_sec = in->lkl_st_ctime;
+	st->st_ctim.tv_nsec = in->st_ctime_nsec;
+}
+
+static int lklfuse_fgetattr(const char *path, struct stat *st,
+			    struct fuse_file_info *fi)
+{
+	long ret;
+	struct lkl_stat lkl_stat;
+
+	ret = lkl_sys_fstat(fi->fh, &lkl_stat);
+	if (ret)
+		return ret;
+
+	lklfuse_xlat_stat(&lkl_stat, st);
+	return 0;
+}
+
+static int lklfuse_getattr(const char *path, struct stat *st)
+{
+	long ret;
+	struct lkl_stat lkl_stat;
+
+	ret = lkl_sys_lstat(path, &lkl_stat);
+	if (ret)
+		return ret;
+
+	lklfuse_xlat_stat(&lkl_stat, st);
+	return 0;
+}
+
+static int lklfuse_readlink(const char *path, char *buf, size_t len)
+{
+	long ret;
+
+	ret = lkl_sys_readlink(path, buf, len);
+	if (ret < 0)
+		return ret;
+
+	if ((size_t)ret == len)
+		ret = len - 1;
+
+	buf[ret] = 0;
+
+	return 0;
+}
+
+static int lklfuse_mknod(const char *path, mode_t mode, dev_t dev)
+{
+	return lkl_sys_mknod(path, mode, dev);
+}
+
+static int lklfuse_mkdir(const char *path, mode_t mode)
+{
+	return lkl_sys_mkdir(path, mode);
+}
+
+static int lklfuse_unlink(const char *path)
+{
+	return lkl_sys_unlink(path);
+}
+
+static int lklfuse_rmdir(const char *path)
+{
+	return lkl_sys_rmdir(path);
+}
+
+static int lklfuse_symlink(const char *oldname, const char *newname)
+{
+	return lkl_sys_symlink(oldname, newname);
+}
+
+
+static int lklfuse_rename(const char *oldname, const char *newname)
+{
+	return lkl_sys_rename(oldname, newname);
+}
+
+static int lklfuse_link(const char *oldname, const char *newname)
+{
+	return lkl_sys_link(oldname, newname);
+}
+
+static int lklfuse_chmod(const char *path, mode_t mode)
+{
+	return lkl_sys_chmod(path, mode);
+}
+
+
+static int lklfuse_chown(const char *path, uid_t uid, gid_t gid)
+{
+	return lkl_sys_fchownat(LKL_AT_FDCWD, path, uid, gid,
+				LKL_AT_SYMLINK_NOFOLLOW);
+}
+
+static int lklfuse_truncate(const char *path, off_t off)
+{
+	return lkl_sys_truncate(path, off);
+}
+
+static int lklfuse_open3(const char *path, bool create, mode_t mode,
+			 struct fuse_file_info *fi)
+{
+	long ret;
+	int flags;
+
+	if ((fi->flags & O_ACCMODE) == O_RDONLY)
+		flags = LKL_O_RDONLY;
+	else if ((fi->flags & O_ACCMODE) == O_WRONLY)
+		flags = LKL_O_WRONLY;
+	else if ((fi->flags & O_ACCMODE) == O_RDWR)
+		flags = LKL_O_RDWR;
+	else
+		return -EINVAL;
+
+	if (create)
+		flags |= LKL_O_CREAT;
+
+	ret = lkl_sys_open(path, flags, mode);
+	if (ret < 0)
+		return ret;
+
+	fi->fh = ret;
+
+	return 0;
+}
+
+static int lklfuse_create(const char *path, mode_t mode,
+			  struct fuse_file_info *fi)
+{
+	return lklfuse_open3(path, true, mode, fi);
+}
+
+static int lklfuse_open(const char *path, struct fuse_file_info *fi)
+{
+	return lklfuse_open3(path, false, 0, fi);
+}
+
+static int lklfuse_read(const char *path, char *buf, size_t size, off_t offset,
+		      struct fuse_file_info *fi)
+{
+	long ret;
+	ssize_t orig_size = size;
+
+	do {
+		ret = lkl_sys_pread64(fi->fh, buf, size, offset);
+		if (ret <= 0)
+			break;
+		size -= ret;
+		offset += ret;
+		buf += ret;
+	} while (size > 0);
+
+	return ret < 0 ? ret : orig_size - (ssize_t)size;
+
+}
+
+static int lklfuse_write(const char *path, const char *buf, size_t size,
+		       off_t offset, struct fuse_file_info *fi)
+{
+	long ret;
+	ssize_t orig_size = size;
+
+	do {
+		ret = lkl_sys_pwrite64(fi->fh, buf, size, offset);
+		if (ret <= 0)
+			break;
+		size -= ret;
+		offset += ret;
+		buf += ret;
+	} while (size > 0);
+
+	return ret < 0 ? ret : orig_size - (ssize_t)size;
+}
+
+
+static int lklfuse_statfs(const char *path, struct statvfs *stat)
+{
+	long ret;
+	struct lkl_statfs lkl_statfs;
+
+	ret = lkl_sys_statfs(path, &lkl_statfs);
+	if (ret < 0)
+		return ret;
+
+	stat->f_bsize = lkl_statfs.f_bsize;
+	stat->f_frsize = lkl_statfs.f_frsize;
+	stat->f_blocks = lkl_statfs.f_blocks;
+	stat->f_bfree = lkl_statfs.f_bfree;
+	stat->f_bavail = lkl_statfs.f_bavail;
+	stat->f_files = lkl_statfs.f_files;
+	stat->f_ffree = lkl_statfs.f_ffree;
+	stat->f_favail = stat->f_ffree;
+	stat->f_fsid = *(unsigned long *)&lkl_statfs.f_fsid.val[0];
+	stat->f_flag = lkl_statfs.f_flags;
+	stat->f_namemax = lkl_statfs.f_namelen;
+
+	return 0;
+}
+
+static int lklfuse_flush(const char *path, struct fuse_file_info *fi)
+{
+	return 0;
+}
+
+static int lklfuse_release(const char *path, struct fuse_file_info *fi)
+{
+	return lkl_sys_close(fi->fh);
+}
+
+static int lklfuse_fsync(const char *path, int datasync,
+		       struct fuse_file_info *fi)
+{
+	if (datasync)
+		return lkl_sys_fdatasync(fi->fh);
+	else
+		return lkl_sys_fsync(fi->fh);
+}
+
+static int lklfuse_setxattr(const char *path, const char *name, const char *val,
+		   size_t size, int flags)
+{
+	return lkl_sys_setxattr(path, name, val, size, flags);
+}
+
+static int lklfuse_getxattr(const char *path, const char *name, char *val,
+			  size_t size)
+{
+	return lkl_sys_getxattr(path, name, val, size);
+}
+
+static int lklfuse_listxattr(const char *path, char *list, size_t size)
+{
+	return lkl_sys_listxattr(path, list, size);
+}
+
+static int lklfuse_removexattr(const char *path, const char *name)
+{
+	return lkl_sys_removexattr(path, name);
+}
+
+static int lklfuse_opendir(const char *path, struct fuse_file_info *fi)
+{
+	struct lkl_dir *dir;
+	int err;
+
+	dir = lkl_opendir(path, &err);
+	if (!dir)
+		return err;
+
+	fi->fh = (uintptr_t)dir;
+
+	return 0;
+}
+
+/** Read directory
+ *
+ * This supersedes the old getdir() interface.  New applications
+ * should use this.
+ *
+ * The filesystem may choose between two modes of operation:
+ *
+ * 1) The readdir implementation ignores the offset parameter, and
+ * passes zero to the filler function's offset.  The filler
+ * function will not return '1' (unless an error happens), so the
+ * whole directory is read in a single readdir operation.  This
+ * works just like the old getdir() method.
+ *
+ * 2) The readdir implementation keeps track of the offsets of the
+ * directory entries.  It uses the offset parameter and always
+ * passes non-zero offset to the filler function.  When the buffer
+ * is full (or an error happens) the filler function will return
+ * '1'.
+ *
+ * Introduced in version 2.3
+ */
+static int lklfuse_readdir(const char *path, void *buf, fuse_fill_dir_t fill,
+			 off_t off, struct fuse_file_info *fi)
+{
+	struct lkl_dir *dir = (struct lkl_dir *)(uintptr_t)fi->fh;
+	struct lkl_linux_dirent64 *de;
+
+	while ((de = lkl_readdir(dir))) {
+		struct stat st = { 0, };
+
+		st.st_ino = de->d_ino;
+		st.st_mode = de->d_type << 12;
+
+		if (fill(buf, de->d_name, &st, 0))
+			break;
+	}
+
+	if (!de)
+		return lkl_errdir(dir);
+
+	return 0;
+}
+
+static int lklfuse_releasedir(const char *path, struct fuse_file_info *fi)
+{
+	struct lkl_dir *dir = (struct lkl_dir *)(uintptr_t)fi->fh;
+
+	return lkl_closedir(dir);
+}
+
+static int lklfuse_fsyncdir(const char *path, int datasync,
+			  struct fuse_file_info *fi)
+{
+	struct lkl_dir *dir = (struct lkl_dir *)(uintptr_t)fi->fh;
+	int fd = lkl_dirfd(dir);
+
+	if (datasync)
+		return lkl_sys_fdatasync(fd);
+	else
+		return lkl_sys_fsync(fd);
+}
+
+static int lklfuse_access(const char *path, int mode)
+{
+	return lkl_sys_access(path, mode);
+}
+
+static int lklfuse_utimens(const char *path, const struct timespec tv[2])
+{
+	struct lkl_timespec ts[2];
+
+	ts[0].tv_sec = tv[0].tv_sec;
+	ts[0].tv_nsec = tv[0].tv_nsec;
+	ts[1].tv_sec = tv[0].tv_sec;
+	ts[1].tv_nsec = tv[0].tv_nsec;
+
+	return lkl_sys_utimensat(-1, path, (struct __lkl__kernel_timespec *)ts,
+				 LKL_AT_SYMLINK_NOFOLLOW);
+}
+
+static int lklfuse_fallocate(const char *path, int mode, off_t offset,
+			     off_t len, struct fuse_file_info *fi)
+{
+	return lkl_sys_fallocate(fi->fh, mode, offset, len);
+}
+
+const struct fuse_operations lklfuse_ops = {
+	.flag_nullpath_ok = 1,
+	.flag_nopath = 1,
+	.flag_utime_omit_ok = 1,
+
+	.getattr = lklfuse_getattr,
+	.readlink = lklfuse_readlink,
+	.mknod = lklfuse_mknod,
+	.mkdir = lklfuse_mkdir,
+	.unlink = lklfuse_unlink,
+	.rmdir = lklfuse_rmdir,
+	.symlink = lklfuse_symlink,
+	.rename = lklfuse_rename,
+	.link = lklfuse_link,
+	.chmod = lklfuse_chmod,
+	.chown = lklfuse_chown,
+	.truncate = lklfuse_truncate,
+	.open = lklfuse_open,
+	.read = lklfuse_read,
+	.write = lklfuse_write,
+	.statfs = lklfuse_statfs,
+	.flush = lklfuse_flush,
+	.release = lklfuse_release,
+	.fsync = lklfuse_fsync,
+	.setxattr = lklfuse_setxattr,
+	.getxattr = lklfuse_getxattr,
+	.listxattr = lklfuse_listxattr,
+	.removexattr = lklfuse_removexattr,
+	.opendir = lklfuse_opendir,
+	.readdir = lklfuse_readdir,
+	.releasedir = lklfuse_releasedir,
+	.fsyncdir = lklfuse_fsyncdir,
+	.access = lklfuse_access,
+	.create = lklfuse_create,
+	.fgetattr = lklfuse_fgetattr,
+	/* .lock, */
+	.utimens = lklfuse_utimens,
+	/* .bmap, */
+	/* .ioctl, */
+	/* .poll, */
+	/* .write_buf, (SG io) */
+	/* .read_buf, (SG io) */
+	/* .flock, */
+	.fallocate = lklfuse_fallocate,
+};
+
+static int start_lkl(void)
+{
+	long ret;
+	char mpoint[32], cmdline[16];
+
+	snprintf(cmdline, sizeof(cmdline), "mem=%dM", lklfuse.mb);
+	ret = lkl_start_kernel(&lkl_host_ops, cmdline);
+	if (ret) {
+		fprintf(stderr, "can't start kernel: %s\n", lkl_strerror(ret));
+		goto out;
+	}
+
+	ret = lkl_mount_dev(lklfuse.disk_id, lklfuse.part, lklfuse.type,
+			    lklfuse.ro ? LKL_MS_RDONLY : 0, lklfuse.opts,
+			    mpoint, sizeof(mpoint));
+
+	if (ret) {
+		fprintf(stderr, "can't mount disk: %s\n", lkl_strerror(ret));
+		goto out_halt;
+	}
+
+	ret = lkl_sys_chroot(mpoint);
+	if (ret) {
+		fprintf(stderr, "can't chdir to %s: %s\n", mpoint,
+			lkl_strerror(ret));
+		goto out_umount;
+	}
+
+	return 0;
+
+out_umount:
+	lkl_umount_dev(lklfuse.disk_id, lklfuse.part, 0, 1000);
+
+out_halt:
+	lkl_sys_halt();
+
+out:
+	return ret;
+}
+
+static void stop_lkl(void)
+{
+	int ret;
+
+	ret = lkl_sys_chdir("/");
+	if (ret)
+		fprintf(stderr, "can't chdir to /: %s\n", lkl_strerror(ret));
+	ret = lkl_sys_umount("/", 0);
+	if (ret)
+		fprintf(stderr, "failed to umount disk: %d: %s\n",
+			lklfuse.disk_id, lkl_strerror(ret));
+	lkl_sys_halt();
+}
+
+int main(int argc, char **argv)
+{
+	struct fuse_args args = FUSE_ARGS_INIT(argc, argv);
+	struct fuse_chan *ch;
+	struct fuse *fuse;
+	struct stat st;
+	char *mnt;
+	int fg, mt, ret;
+
+	if (fuse_opt_parse(&args, &lklfuse, lklfuse_opts, lklfuse_opt_proc))
+		return 1;
+
+	if (!lklfuse.file || !lklfuse.type) {
+		fprintf(stderr, "no file or filesystem type specified\n");
+		return 1;
+	}
+
+	if (fuse_parse_cmdline(&args, &mnt, &mt, &fg))
+		return 1;
+
+	ret = stat(mnt, &st);
+	if (ret) {
+		perror(mnt);
+		goto out_free;
+	}
+
+	ret = open(lklfuse.file, lklfuse.ro ? O_RDONLY : O_RDWR);
+	if (ret < 0) {
+		perror(lklfuse.file);
+		goto out_free;
+	}
+
+	lklfuse.disk.fd = ret;
+
+	ret = lkl_disk_add(&lklfuse.disk);
+	if (ret < 0) {
+		fprintf(stderr, "can't add disk: %s\n", lkl_strerror(ret));
+		goto out_close_disk;
+	}
+
+	lklfuse.disk_id = ret;
+
+	ch = fuse_mount(mnt, &args);
+	if (!ch) {
+		ret = -1;
+		goto out_close_disk;
+	}
+
+	fuse = fuse_new(ch, &args, &lklfuse_ops, sizeof(lklfuse_ops), NULL);
+	if (!fuse) {
+		ret = -1;
+		goto out_fuse_unmount;
+	}
+
+	fuse_opt_free_args(&args);
+
+	if (fuse_daemonize(fg) ||
+	    fuse_set_signal_handlers(fuse_get_session(fuse))) {
+		ret = -1;
+		goto out_fuse_destroy;
+	}
+
+	ret = start_lkl();
+	if (ret) {
+		ret = -1;
+		goto out_remove_signals;
+	}
+
+	if (mt)
+		fprintf(stderr, "warning: multithreaded mode not supported\n");
+
+	ret = fuse_loop(fuse);
+
+	stop_lkl();
+
+out_remove_signals:
+	fuse_remove_signal_handlers(fuse_get_session(fuse));
+
+out_fuse_unmount:
+	if (ch)
+		fuse_unmount(mnt, ch);
+
+out_fuse_destroy:
+	if (fuse)
+		fuse_destroy(fuse);
+
+out_close_disk:
+	close(lklfuse.disk.fd);
+
+out_free:
+	free(mnt);
+
+	return ret < 0 ? 1 : 0;
+}
diff --git a/tools/lkl/tests/lklfuse.sh b/tools/lkl/tests/lklfuse.sh
new file mode 100755
index 000000000000..7f35dd53fc4e
--- /dev/null
+++ b/tools/lkl/tests/lklfuse.sh
@@ -0,0 +1,110 @@
+#!/usr/bin/env bash
+# SPDX-License-Identifier: GPL-2.0
+
+script_dir=$(cd $(dirname ${BASH_SOURCE:-$0}); pwd)
+
+source $script_dir/test.sh
+
+cleanup()
+{
+    set -e
+
+    sleep 1
+    fusermount -u $dir
+    rm $file
+    rmdir $dir
+}
+
+
+# $1 - disk image
+# $2 - fstype
+function prepfs()
+{
+    set -e
+
+    dd if=/dev/zero of=$1 bs=1024 count=102400
+
+    yes | mkfs.$2 $1
+}
+
+# $1 - disk image
+# $2 - mount point
+# $3 - filesystem type
+lklfuse_mount()
+{
+    ${script_dir}/../lklfuse $1 $2 -o type=$3
+}
+
+# $1 - mount point
+lklfuse_basic()
+{
+    set -e
+
+    cd $1
+    touch a
+    if ! [ -e ]; then exit 1; fi
+    rm a
+    mkdir a
+    if ! [ -d ]; then exit 1; fi
+    rmdir a
+}
+
+# $1 - dir
+# $2 - filesystem type
+lklfuse_stressng()
+{
+    set -e
+
+    if [ -z $(which stress-ng) ]; then
+        echo "missing stress-ng"
+        return $TEST_SKIP
+    fi
+
+    cd $1
+
+    if [ "$2" = "vfat" ]; then
+        exclude="chmod,filename,link,mknod,symlink,xattr"
+    fi
+
+    stress-ng --class filesystem --all 0 --timeout 10 \
+	      --exclude fiemap,$exclude --fallocate-bytes 10m \
+	      --sync-file-bytes 10m
+}
+
+if [ "$1" = "-t" ]; then
+    shift
+    fstype=$1
+    shift
+fi
+
+if [ -z "$fstype" ]; then
+    fstype="ext4"
+fi
+
+if ! [ -x $script_dir/../lklfuse ]; then
+    lkl_test_plan 0 "lklfuse.sh $fstype"
+    echo "lklfuse not available"
+    exit 0
+fi
+
+if ! [ -e /dev/fuse ]; then
+    lkl_test_plan 0 "lklfuse.sh $fstype"
+    echo "/dev/fuse not available"
+    exit 0
+fi
+
+
+file=`mktemp`
+dir=`mktemp -d`
+
+trap cleanup EXIT
+
+lkl_test_plan 4 "lklfuse $fstype"
+
+lkl_test_run 1 prepfs $file $fstype
+lkl_test_run 2 lklfuse_mount $file $dir $fstype
+lkl_test_run 3 lklfuse_basic $dir
+# stress-ng returns 2 with no apparent failures so skip it for now
+#lkl_test_run 4 lklfuse_stressng $dir $fstype
+trap : EXIT
+lkl_test_run 4 cleanup
-- 
2.20.1 (Apple Git-117)


_______________________________________________
linux-um mailing list
linux-um@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-um