d_path() and overlay fs

d_path() and overlay fs

[Karel Zak]

 Hi,

 kernel-4.0.0-0.rc4.git1.3.fc23 (I have nothing more recent right now,
 sorry it the problem is already fixed).
 
 It seems like d_path() kernel function returns nonsenses for overlay fs.

 Example:

   # mount -t overlay overlay -olowerdir=/mnt/low,upperdir=/mnt/up/data,workdir=/mnt/up/work /mnt/merged
   # dd if=/dev/zero of=/mnt/merged/image count=1 bs=1MiB

   # losetup -f /mnt/merged/image

   # cat /sys/block/loop0/loop/backing_file 
   /image


 ...should be /mnt/merged/image.

 drivers/block/loop.c: loop_attr_backing_file_show() uses d_path()

 It seems like serious problem because kernel uses d_path() (and
 friends) on many places to generate path strings for userspace in
 /proc and /sys and people already use overlay fs in containers.
 
 Another example:
 
   # cat /proc/swaps 
   Filename               Type        Size    Used    Priority
   /dev/sda6              partition   8191996 258308  -1
   /image                 file        1020    0       -2


    Karel

-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com

Re: d_path() and overlay fs

[Al Viro]

On Fri, Mar 20, 2015 at 02:29:14PM +0100, Karel Zak wrote:

>  It seems like serious problem because kernel uses d_path() (and
>  friends) on many places to generate path strings for userspace in
>  /proc and /sys and people already use overlay fs in containers.

It's not d_path(), it's file->f_path being wrong to start with.  Same
problem as what dhowells is fighting wrt LSM shite.

Re: d_path() and overlay fs

[Josh Boyer]

On Fri, Mar 20, 2015 at 9:29 AM, Karel Zak <kzak@redhat.com> wrote:
>
>  Hi,
>
>  kernel-4.0.0-0.rc4.git1.3.fc23 (I have nothing more recent right now,
>  sorry it the problem is already fixed).

That equates to Linux v4.0-rc4-88-g7b09ac704bac upstream.

Fedora has Linux v4.0-rc4-199-gb314acaccd7e queued to build today,
which does include an overlayfs fix pull but I have no idea if it
solves the issue.

josh

>
>  It seems like d_path() kernel function returns nonsenses for overlay fs.
>
>  Example:
>
>    # mount -t overlay overlay -olowerdir=/mnt/low,upperdir=/mnt/up/data,workdir=/mnt/up/work /mnt/merged
>    # dd if=/dev/zero of=/mnt/merged/image count=1 bs=1MiB
>
>    # losetup -f /mnt/merged/image
>
>    # cat /sys/block/loop0/loop/backing_file
>    /image
>
>
>  ...should be /mnt/merged/image.
>
>  drivers/block/loop.c: loop_attr_backing_file_show() uses d_path()
>
>  It seems like serious problem because kernel uses d_path() (and
>  friends) on many places to generate path strings for userspace in
>  /proc and /sys and people already use overlay fs in containers.
>
>  Another example:
>
>    # cat /proc/swaps
>    Filename               Type        Size    Used    Priority
>    /dev/sda6              partition   8191996 258308  -1
>    /image                 file        1020    0       -2
>
>
>     Karel
>
> --
>  Karel Zak  <kzak@redhat.com>
>  http://karelzak.blogspot.com
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Re: d_path() and overlay fs

[Miklos Szeredi]

On Fri, Mar 20, 2015 at 09:41:25AM -0400, Josh Boyer wrote:
> On Fri, Mar 20, 2015 at 9:29 AM, Karel Zak <kzak@redhat.com> wrote:
> >
> >  Hi,
> >
> >  kernel-4.0.0-0.rc4.git1.3.fc23 (I have nothing more recent right now,
> >  sorry it the problem is already fixed).
> 
> That equates to Linux v4.0-rc4-88-g7b09ac704bac upstream.
> 
> Fedora has Linux v4.0-rc4-199-gb314acaccd7e queued to build today,
> which does include an overlayfs fix pull but I have no idea if it
> solves the issue.

It does not.

However the following patch should.  It's not much tested and I understand that
it doesn't solve the whole problem that dhowells is working on (selinux in
particular)

But it does take care of the majority of f_path users that actually want the
covering path.

Comments, testing welcome.

Thanks,
Miklos

---
 arch/arc/kernel/troubleshoot.c               |   10 ++------
 arch/blackfin/kernel/trace.c                 |    2 -
 arch/tile/kernel/stack.c                     |    2 -
 arch/tile/mm/elf.c                           |    2 -
 drivers/block/loop.c                         |    2 -
 drivers/md/bitmap.c                          |    4 +--
 drivers/md/md.c                              |    2 -
 drivers/usb/gadget/function/f_mass_storage.c |    2 -
 drivers/usb/gadget/function/storage_common.c |    2 -
 fs/binfmt_elf.c                              |    4 +--
 fs/coredump.c                                |    2 -
 fs/ext4/super.c                              |    2 -
 fs/file_table.c                              |    1 
 fs/open.c                                    |   32 +++++++++++++++++++++++----
 fs/proc/base.c                               |    6 ++---
 fs/proc/fd.c                                 |    9 +++++--
 fs/proc/nommu.c                              |    2 -
 fs/proc/task_mmu.c                           |    4 +--
 fs/proc/task_nommu.c                         |    2 -
 fs/seq_file.c                                |   17 ++++++++++++++
 include/linux/fs.h                           |    4 +++
 include/linux/seq_file.h                     |    1 
 kernel/events/core.c                         |    2 -
 mm/memory.c                                  |    2 -
 mm/swapfile.c                                |    2 -
 25 files changed, 83 insertions(+), 37 deletions(-)

--- a/fs/file_table.c
+++ b/fs/file_table.c
@@ -219,6 +219,7 @@ static void __fput(struct file *file)
 		put_write_access(inode);
 		__mnt_drop_write(mnt);
 	}
+	path_put(&file->f_covering_path);
 	file->f_path.dentry = NULL;
 	file->f_path.mnt = NULL;
 	file->f_inode = NULL;
--- a/fs/open.c
+++ b/fs/open.c
@@ -849,6 +849,24 @@ struct file *dentry_open(const struct pa
 }
 EXPORT_SYMBOL(dentry_open);
 
+void f_covering_path(struct file *filp, struct path *path)
+{
+	if (filp->f_covering_path.dentry)
+		*path = filp->f_covering_path;
+	else
+		*path = filp->f_path;
+}
+EXPORT_SYMBOL(f_covering_path);
+
+char *file_path(struct file *filp, char *buf, int buflen)
+{
+	struct path path;
+
+	f_covering_path(filp, &path);
+	return d_path(&path, buf, buflen);
+}
+EXPORT_SYMBOL(file_path);
+
 /**
  * vfs_open - open the file at the given path
  * @path: path to open
@@ -859,13 +877,19 @@ int vfs_open(const struct path *path, st
 	     const struct cred *cred)
 {
 	struct inode *inode = path->dentry->d_inode;
+	int err;
 
-	if (inode->i_op->dentry_open)
-		return inode->i_op->dentry_open(path->dentry, filp, cred);
-	else {
+	if (inode->i_op->dentry_open) {
+		err = inode->i_op->dentry_open(path->dentry, filp, cred);
+		if (!err && filp->f_path.dentry != path->dentry) {
+			filp->f_covering_path = *path;
+			path_get(&filp->f_covering_path);
+		}
+	} else {
 		filp->f_path = *path;
-		return do_dentry_open(filp, NULL, cred);
+		err = do_dentry_open(filp, NULL, cred);
 	}
+	return err;
 }
 EXPORT_SYMBOL(vfs_open);
 
--- a/fs/proc/base.c
+++ b/fs/proc/base.c
@@ -1363,8 +1363,8 @@ static int proc_exe_link(struct dentry *
 	exe_file = get_mm_exe_file(mm);
 	mmput(mm);
 	if (exe_file) {
-		*exe_path = exe_file->f_path;
-		path_get(&exe_file->f_path);
+		f_covering_path(exe_file, exe_path);
+		path_get(exe_path);
 		fput(exe_file);
 		return 0;
 	} else
@@ -1735,7 +1735,7 @@ static int proc_map_files_get_link(struc
 	down_read(&mm->mmap_sem);
 	vma = find_exact_vma(mm, vm_start, vm_end);
 	if (vma && vma->vm_file) {
-		*path = vma->vm_file->f_path;
+		f_covering_path(vma->vm_file, path);
 		path_get(path);
 		rc = 0;
 	}
--- a/fs/proc/fd.c
+++ b/fs/proc/fd.c
@@ -49,9 +49,12 @@ static int seq_show(struct seq_file *m,
 	}
 
 	if (!ret) {
+		struct path path;
+
+		f_covering_path(file, &path);
 		seq_printf(m, "pos:\t%lli\nflags:\t0%o\nmnt_id:\t%i\n",
 			   (long long)file->f_pos, f_flags,
-			   real_mount(file->f_path.mnt)->mnt_id);
+			   real_mount(path.mnt)->mnt_id);
 		if (file->f_op->show_fdinfo)
 			file->f_op->show_fdinfo(m, file);
 		ret = seq_has_overflowed(m);
@@ -157,8 +160,8 @@ static int proc_fd_link(struct dentry *d
 		spin_lock(&files->file_lock);
 		fd_file = fcheck_files(files, fd);
 		if (fd_file) {
-			*path = fd_file->f_path;
-			path_get(&fd_file->f_path);
+			f_covering_path(fd_file, path);
+			path_get(path);
 			ret = 0;
 		}
 		spin_unlock(&files->file_lock);
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -833,6 +833,7 @@ struct file {
 	struct list_head	f_tfile_llink;
 #endif /* #ifdef CONFIG_EPOLL */
 	struct address_space	*f_mapping;
+	struct path             f_covering_path;
 } __attribute__((aligned(4)));	/* lest something weird decides that 2 is OK */
 
 struct file_handle {
@@ -2466,6 +2467,9 @@ extern struct file * open_exec(const cha
 extern int is_subdir(struct dentry *, struct dentry *);
 extern int path_is_under(struct path *, struct path *);
 
+extern char *file_path(struct file *, char *, int);
+extern void f_covering_path(struct file *, struct path *);
+
 #include <linux/err.h>
 
 /* needed for stackable file system support */
--- a/drivers/block/loop.c
+++ b/drivers/block/loop.c
@@ -620,7 +620,7 @@ static ssize_t loop_attr_backing_file_sh
 
 	spin_lock_irq(&lo->lo_lock);
 	if (lo->lo_backing_file)
-		p = d_path(&lo->lo_backing_file->f_path, buf, PAGE_SIZE - 1);
+		p = file_path(lo->lo_backing_file, buf, PAGE_SIZE - 1);
 	spin_unlock_irq(&lo->lo_lock);
 
 	if (IS_ERR_OR_NULL(p))
--- a/include/linux/seq_file.h
+++ b/include/linux/seq_file.h
@@ -123,6 +123,7 @@ __printf(2, 3) int seq_printf(struct seq
 __printf(2, 0) int seq_vprintf(struct seq_file *, const char *, va_list args);
 
 int seq_path(struct seq_file *, const struct path *, const char *);
+int seq_file_path(struct seq_file *, struct file *, const char *);
 int seq_dentry(struct seq_file *, struct dentry *, const char *);
 int seq_path_root(struct seq_file *m, const struct path *path,
 		  const struct path *root, const char *esc);
--- a/fs/seq_file.c
+++ b/fs/seq_file.c
@@ -487,6 +487,23 @@ int seq_path(struct seq_file *m, const s
 }
 EXPORT_SYMBOL(seq_path);
 
+/**
+ * seq_file_path - seq_file interface to print a pathname of a file
+ * @m: the seq_file handle
+ * @file: the struct file to print
+ * @esc: set of characters to escape in the output
+ *
+ * return the absolute path to the file.
+ */
+int seq_file_path(struct seq_file *m, struct file *file, const char *esc)
+{
+	struct path path;
+
+	f_covering_path(file, &path);
+	return seq_path(m, &path, esc);
+}
+EXPORT_SYMBOL(seq_file_path);
+
 /*
  * Same as seq_path, but relative to supplied root.
  */
--- a/fs/proc/nommu.c
+++ b/fs/proc/nommu.c
@@ -64,7 +64,7 @@ static int nommu_region_show(struct seq_
 
 	if (file) {
 		seq_pad(m, ' ');
-		seq_path(m, &file->f_path, "");
+		seq_file_path(m, file, "");
 	}
 
 	seq_putc(m, '\n');
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c
@@ -310,7 +310,7 @@ show_map_vma(struct seq_file *m, struct
 	 */
 	if (file) {
 		seq_pad(m, ' ');
-		seq_path(m, &file->f_path, "\n");
+		seq_file_path(m, file, "\n");
 		goto done;
 	}
 
@@ -1509,7 +1509,7 @@ static int show_numa_map(struct seq_file
 
 	if (file) {
 		seq_puts(m, " file=");
-		seq_path(m, &file->f_path, "\n\t= ");
+		seq_file_path(m, file, "\n\t= ");
 	} else if (vma->vm_start <= mm->brk && vma->vm_end >= mm->start_brk) {
 		seq_puts(m, " heap");
 	} else {
--- a/fs/proc/task_nommu.c
+++ b/fs/proc/task_nommu.c
@@ -180,7 +180,7 @@ static int nommu_vma_show(struct seq_fil
 
 	if (file) {
 		seq_pad(m, ' ');
-		seq_path(m, &file->f_path, "");
+		seq_file_path(m, file, "");
 	} else if (mm) {
 		pid_t tid = pid_of_stack(priv, vma, is_pid);
 
--- a/mm/swapfile.c
+++ b/mm/swapfile.c
@@ -2032,7 +2032,7 @@ static int swap_show(struct seq_file *sw
 	}
 
 	file = si->swap_file;
-	len = seq_path(swap, &file->f_path, " \t\n\\");
+	len = seq_file_path(swap, file, " \t\n\\");
 	seq_printf(swap, "%*s%s\t%u\t%u\t%d\n",
 			len < 40 ? 40 - len : 1, " ",
 			S_ISBLK(file_inode(file)->i_mode) ?
--- a/arch/arc/kernel/troubleshoot.c
+++ b/arch/arc/kernel/troubleshoot.c
@@ -67,15 +67,12 @@ void print_task_path_n_nm(struct task_st
 	mmput(mm);
 
 	if (exe_file) {
-		path = exe_file->f_path;
-		path_get(&exe_file->f_path);
+		path_nm = file_path(exe_file, buf, 255);
 		fput(exe_file);
-		path_nm = d_path(&path, buf, 255);
-		path_put(&path);
 	}
 
 done:
-	pr_info("Path: %s\n", path_nm);
+	pr_info("Path: %s\n", !IS_ERR(path_nm) ? path_nm : "?");
 }
 EXPORT_SYMBOL(print_task_path_n_nm);
 
@@ -100,8 +97,7 @@ static void show_faulting_vma(unsigned l
 	if (vma && (vma->vm_start <= address)) {
 		struct file *file = vma->vm_file;
 		if (file) {
-			struct path *path = &file->f_path;
-			nm = d_path(path, buf, PAGE_SIZE - 1);
+			nm = file_path(file, buf, PAGE_SIZE - 1);
 			inode = file_inode(vma->vm_file);
 			dev = inode->i_sb->s_dev;
 			ino = inode->i_ino;
--- a/arch/blackfin/kernel/trace.c
+++ b/arch/blackfin/kernel/trace.c
@@ -136,7 +136,7 @@ void decode_address(char *buf, unsigned
 				struct file *file = vma->vm_file;
 
 				if (file) {
-					char *d_name = d_path(&file->f_path, _tmpbuf,
+					char *d_name = file_path(file, _tmpbuf,
 						      sizeof(_tmpbuf));
 					if (!IS_ERR(d_name))
 						name = d_name;
--- a/arch/tile/kernel/stack.c
+++ b/arch/tile/kernel/stack.c
@@ -333,7 +333,7 @@ static void describe_addr(struct KBacktr
 	}
 
 	if (vma->vm_file) {
-		p = d_path(&vma->vm_file->f_path, buf, bufsize);
+		p = file_path(vma->vm_file, buf, bufsize);
 		if (IS_ERR(p))
 			p = "?";
 		name = kbasename(p);
--- a/arch/tile/mm/elf.c
+++ b/arch/tile/mm/elf.c
@@ -59,7 +59,7 @@ static int notify_exec(struct mm_struct
 	if (buf == NULL)
 		return 0;
 
-	path = d_path(&mm->exe_file->f_path, buf, PAGE_SIZE);
+	path = file_path(mm->exe_file, buf, PAGE_SIZE);
 	if (IS_ERR(path)) {
 		free_page((unsigned long)buf);
 		return 0;
--- a/drivers/md/bitmap.c
+++ b/drivers/md/bitmap.c
@@ -785,7 +785,7 @@ static void bitmap_file_kick(struct bitm
 		if (bitmap->storage.file) {
 			path = kmalloc(PAGE_SIZE, GFP_KERNEL);
 			if (path)
-				ptr = d_path(&bitmap->storage.file->f_path,
+				ptr = file_path(bitmap->storage.file,
 					     path, PAGE_SIZE);
 
 			printk(KERN_ALERT
@@ -1786,7 +1786,7 @@ void bitmap_status(struct seq_file *seq,
 		   chunk_kb ? "KB" : "B");
 	if (bitmap->storage.file) {
 		seq_printf(seq, ", file: ");
-		seq_path(seq, &bitmap->storage.file->f_path, " \t\n");
+		seq_file_path(seq, bitmap->storage.file, " \t\n");
 	}
 
 	seq_printf(seq, "\n");
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -5660,7 +5660,7 @@ static int get_bitmap_file(struct mddev
 	/* bitmap disabled, zero the first byte and copy out */
 	if (!mddev->bitmap_info.file)
 		file->pathname[0] = '\0';
-	else if ((ptr = d_path(&mddev->bitmap_info.file->f_path,
+	else if ((ptr = file_path(mddev->bitmap_info.file,
 			       file->pathname, sizeof(file->pathname))),
 		 IS_ERR(ptr))
 		err = PTR_ERR(ptr);
--- a/drivers/usb/gadget/function/f_mass_storage.c
+++ b/drivers/usb/gadget/function/f_mass_storage.c
@@ -2973,7 +2973,7 @@ int fsg_common_create_lun(struct fsg_com
 	if (fsg_lun_is_open(lun)) {
 		p = "(error)";
 		if (pathbuf) {
-			p = d_path(&lun->filp->f_path, pathbuf, PATH_MAX);
+			p = file_path(lun->filp, pathbuf, PATH_MAX);
 			if (IS_ERR(p))
 				p = "(error)";
 		}
--- a/drivers/usb/gadget/function/storage_common.c
+++ b/drivers/usb/gadget/function/storage_common.c
@@ -341,7 +341,7 @@ ssize_t fsg_show_file(struct fsg_lun *cu
 
 	down_read(filesem);
 	if (fsg_lun_is_open(curlun)) {	/* Get the complete pathname */
-		p = d_path(&curlun->filp->f_path, buf, PAGE_SIZE - 1);
+		p = file_path(curlun->filp, buf, PAGE_SIZE - 1);
 		if (IS_ERR(p))
 			rc = PTR_ERR(p);
 		else {
--- a/fs/binfmt_elf.c
+++ b/fs/binfmt_elf.c
@@ -1535,7 +1535,7 @@ static int fill_files_note(struct memelf
 		file = vma->vm_file;
 		if (!file)
 			continue;
-		filename = d_path(&file->f_path, name_curpos, remaining);
+		filename = file_path(file, name_curpos, remaining);
 		if (IS_ERR(filename)) {
 			if (PTR_ERR(filename) == -ENAMETOOLONG) {
 				vfree(data);
@@ -1545,7 +1545,7 @@ static int fill_files_note(struct memelf
 			continue;
 		}
 
-		/* d_path() fills at the end, move name down */
+		/* file_path() fills at the end, move name down */
 		/* n = strlen(filename) + 1: */
 		n = (name_curpos + remaining) - filename;
 		remaining = filename - name_curpos;
--- a/fs/coredump.c
+++ b/fs/coredump.c
@@ -138,7 +138,7 @@ static int cn_print_exe_file(struct core
 		goto put_exe_file;
 	}
 
-	path = d_path(&exe_file->f_path, pathbuf, PATH_MAX);
+	path = file_path(exe_file, pathbuf, PATH_MAX);
 	if (IS_ERR(path)) {
 		ret = PTR_ERR(path);
 		goto free_buf;
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -466,7 +466,7 @@ void __ext4_error_file(struct file *file
 	es = EXT4_SB(inode->i_sb)->s_es;
 	es->s_last_error_ino = cpu_to_le32(inode->i_ino);
 	if (ext4_error_ratelimit(inode->i_sb)) {
-		path = d_path(&(file->f_path), pathname, sizeof(pathname));
+		path = file_path(file, pathname, sizeof(pathname));
 		if (IS_ERR(path))
 			path = "(unknown)";
 		va_start(args, fmt);
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -5594,7 +5594,7 @@ static void perf_event_mmap_event(struct
 		 * need to add enough zero bytes after the string to handle
 		 * the 64bit alignment we do later.
 		 */
-		name = d_path(&file->f_path, buf, PATH_MAX - sizeof(u64));
+		name = file_path(file, buf, PATH_MAX - sizeof(u64));
 		if (IS_ERR(name)) {
 			name = "//toolong";
 			goto cpy_name;
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -3627,7 +3627,7 @@ void print_vma_addr(char *prefix, unsign
 		if (buf) {
 			char *p;
 
-			p = d_path(&f->f_path, buf, PAGE_SIZE);
+			p = file_path(f, buf, PAGE_SIZE);
 			if (IS_ERR(p))
 				p = "?";
 			printk("%s%s[%lx+%lx]", prefix, kbasename(p),

Re: d_path() and overlay fs

[Al Viro]

On Fri, Mar 20, 2015 at 05:01:23PM +0100, Miklos Szeredi wrote:

> But it does take care of the majority of f_path users that actually want the
> covering path.

Bloody bad idea, IMO.  I have no objections against adding _helpers_ from
that patch (seq_file_path(), etc.), but I really don't like adding that
second struct path there.  And it still doesn't fix the issue with
LSM, etc., so we'll _still_ need to fix it sane way.

Re: d_path() and overlay fs

[Miklos Szeredi]

On Fri, Mar 20, 2015 at 04:25:58PM +0000, Al Viro wrote:
> On Fri, Mar 20, 2015 at 05:01:23PM +0100, Miklos Szeredi wrote:
> 
> > But it does take care of the majority of f_path users that actually want the
> > covering path.
> 
> Bloody bad idea, IMO.  I have no objections against adding _helpers_ from
> that patch (seq_file_path(), etc.), but I really don't like adding that
> second struct path there.  And it still doesn't fix the issue with
> LSM, etc., so we'll _still_ need to fix it sane way.

Obviously getting rid of the extra path would be good.  But we still have lots
of f_path.dentry in filesystems and we need to start with that.

struct dentry *file_dentry(struct file *) ?  Implemented how?  Rename f_inode to
f_dentry and reimplement file_inode() based on that.

BTW, since nobody is accessing ->f_covering_path directly except the single
f_covering_path() helper, it would be extremely easy to get rid of it later.
That's why I posted this patch, I think it's simple enough to get it into v4.0
which would fix the majority of cases that people complain about.

The thing could even be made dependent on CONFIG_OVERLAY_FS if the addition
actually increases the footprint of struct file (I haven't checked).

Thanks,
Miklos

Re: d_path() and overlay fs

[Miklos Szeredi]

On Fri, Mar 20, 2015 at 5:53 PM, Miklos Szeredi <miklos@szeredi.hu> wrote:
> On Fri, Mar 20, 2015 at 04:25:58PM +0000, Al Viro wrote:
>> On Fri, Mar 20, 2015 at 05:01:23PM +0100, Miklos Szeredi wrote:
>>
>> > But it does take care of the majority of f_path users that actually want the
>> > covering path.
>>
>> Bloody bad idea, IMO.  I have no objections against adding _helpers_ from
>> that patch (seq_file_path(), etc.), but I really don't like adding that
>> second struct path there.  And it still doesn't fix the issue with
>> LSM, etc., so we'll _still_ need to fix it sane way.
>
> Obviously getting rid of the extra path would be good.  But we still have lots
> of f_path.dentry in filesystems and we need to start with that.

So we'd need the following helpers:

file_path() - d_path() but gets file
seq_file_path() - seq_path() but gets file
f_covering_path() - other uses of ->f_path that don't want the backing
filesystem's dentry
file_dentry() - uses of f_path.dentry that want the backing filesystem's dentry
file_mount() - uses of f_path.mnt

Hopefully that would cover all f_path uses.

I'd really not want to keep f_path with that name, because it would
just breed new f_path.dentry uses in filesystems.  Better no direct
f_path references at all.

Thanks,
Miklos

Re: d_path() and overlay fs

[sangeetha Busangari]

Hi,
I am new to Linux kernel programming. i am trying to print current working
directory of process which are running under containers.
i have written code in kernel as below to get pwd
*********************************************************************
struct path pwd;
	char *buf,*cwd;
	uid_t process_uid = (current->cred)->uid.val;
	pwd = current->fs->pwd;
 	path_get(&pwd);//it gets the reference to path. which is argument to
d_path
 	if((buf = (char *)kmalloc(100*sizeof(char),GFP_KERNEL))!=NULL) 
	{       //GFP_KERNEL is a flag specifies the behavior of the memory
allocator
		cwd = d_path(&pwd,buf,100*sizeof(char)); // returns pointer to buffer
printk("process name %s (pid = %d)pwd= %s and (uid=%u)\n",current->comm,
current->pid,cwd,process_uid);
	}
	//printk("check pwd\n");
	kfree(buf); 
**************************************************************************
it is printing pwd correctly in host machine. but unable to start docker
containers

root@sangeetha:/var/lib/docker# docker start 58e9277a7e8c
Error response from daemon: Cannot start container 58e9277a7e8c: [8] System
error: failed to add interface vethebc8876 to sandbox: failed in prefunc:
failed to set namespace on link "vethebc8876": invalid argument
Error: failed to start containers: [58e9277a7e8c]

if i comment d_path line, then i am able to start containers.can you please
help to understand the problem and how to resolve it?

docker version:
Client:
 Version:      1.9.1
 API version:  1.21
 Go version:   go1.4.2
 Git commit:   a34a1d5
 Built:        Fri Nov 20 13:12:04 UTC 2015
 OS/Arch:      linux/amd64

Server:
 Version:      1.9.1
 API version:  1.21
 Go version:   go1.4.2
 Git commit:   a34a1d5
 Built:        Fri Nov 20 13:12:04 UTC 2015
 OS/Arch:      linux/amd64
*******************************************
uname -a
Linux sangeetha 3.14.57 #56 SMP Wed Jan 6 09:37:22 IST 2016 x86_64 x86_64
x86_64 GNU/Linux
**************************************************8
in logs

 process name docker (pid = 3754)pwd= /var/lib/docker and (uid=0)
[ 1743.097731] docker0: port 1(vethcf36114) entered forwarding state
[ 1743.097744] docker0: port 1(vethcf36114) entered forwarding state
[ 1743.100084] docker0: port 1(vethcf36114) entered disabled state
[ 1743.271360] process name docker (pid = 1667)pwd= / and (uid=0)
[ 1743.276427] process name docker (pid = 1507)pwd= / and (uid=0)
[ 1743.496652] docker0: port 1(vethcf36114) entered disabled state
[ 1743.497330] docker0: port 1(vethcf36114) entered disabled state
[ 1799.635904] process name bash (pid = 2894)pwd= /var/lib/docker and
(uid=0)
[ 2543.229052] process name bash (pid = 2894)pwd= /var/lib/docker and
(uid=0)
**************************************************************************
devicemapper fs is used for containers. ubuntu 14.04 os

Thanks in advance.





--
View this message in context: http://linux-kernel.2935.n7.nabble.com/d-path-and-overlay-fs-tp1068779p1273362.html
Sent from the Linux Kernel mailing list archive at Nabble.com.