Re: [PATCH v3 1/5] add metadata_incore ioctl in vfs

On Thu, 2011-01-20 at 04:41 +0800, Andrew Morton wrote:

On Wed, 19 Jan 2011 09:15:18 +0800
Shaohua Li [off-list ref] wrote:

quoted

Subject: add metadata_incore ioctl in vfs

Add an ioctl to dump filesystem's metadata in memory in vfs. Userspace collects
such info and uses it to do metadata readahead.
Filesystem can hook to super_operations.metadata_incore to get metadata in
specific approach. Next patch will give an example how to implement
.metadata_incore in btrfs.

...
 
 /*
+ * Copy info about metadata in memory to userspace
+ * Returns:
+ * = 1, one metadata range copied to userspace
+ * = 0, no more metadata
+ * < 0, error
+ */
+static int ioctl_metadata_incore(struct file *filp, void __user *argp)
+{
+	struct super_block *sb = filp->f_path.dentry->d_inode->i_sb;
+	struct metadata_incore_args args;
+	loff_t offset;
+	ssize_t size;
+
+	if (!sb->s_op->metadata_incore)
+		return -EINVAL;
+
+	if (copy_from_user(&args, argp, sizeof(args)))
+		return -EFAULT;
+
+	/* we check metadata info in page unit */
+	if (args.offset & ~PAGE_CACHE_MASK)
+		return -EINVAL;
+
+	offset = args.offset;
+
+	if (sb->s_op->metadata_incore(sb, &offset, &size) < 0)
+		return 0;
+
+	args.address = offset;
+	args.size = size;
+	args.unused = 0;
+
+	if (copy_to_user(argp, &args, sizeof(args)))
+		return -EFAULT;
+
+	return 1;
+}

So userspace opens any file on the fs and runs this ioctl against it?

That's a pretty awkward interface - we're doing an fs-wide operation
but the fs is identified by a single file which happens to live on that
fs.  For example, this precludes a future extension whereby userspace
can query the incore metadata for a particular file.  The statfs
syscall sucks in the same manner.

I don't know if this is worth addressing.  Perhaps require that the
filp refers to the root of the fs?

I didn't see why this is needed, but I can limit the fip to the root of
the fs.

Also, is this a privileged operation?  If not, then that might be a
problem - could it be used by unprivileged users to work out which
files have been opened recently or something like that?

it's harmless even a unprivileged user uses it. I don't think
unprivileged user can decode the data returned from the ioctl.

Thanks,
Shaohua

On Thu, 20 Jan 2011 10:30:47 +0800 Shaohua Li [off-list ref] wrote:

quoted

I don't know if this is worth addressing.  Perhaps require that the
filp refers to the root of the fs?

I didn't see why this is needed, but I can limit the fip to the root of
the fs.

I don't think it matters much either.  The only problem I can see is if
we were to later try to extend the ioctl into a per-file thing.

quoted

Also, is this a privileged operation?  If not, then that might be a
problem - could it be used by unprivileged users to work out which
files have been opened recently or something like that?

it's harmless even a unprivileged user uses it. I don't think
unprivileged user can decode the data returned from the ioctl.

um.

Well, by doing a before-and-after thing I can use this ioctl to work
out what metadata blocks are used when someone reads
/my/super/secret-directory/foo.  Then I can write a program which sits
there waiting until someone else reads /my/super/secret-directory/foo. 
Then I can use that information to start WWIII or something.

I dunno, strange things happen.  Unless there's a good *need* to make
this available to unprivileged users then we should not do so.

On Thu, 2011-01-20 at 10:42 +0800, Andrew Morton wrote:

On Thu, 20 Jan 2011 10:30:47 +0800 Shaohua Li [off-list ref] wrote:

quoted

quoted

I don't know if this is worth addressing.  Perhaps require that the
filp refers to the root of the fs?

I didn't see why this is needed, but I can limit the fip to the root of
the fs.

I don't think it matters much either.  The only problem I can see is if
we were to later try to extend the ioctl into a per-file thing.

since we return page range, a metadata page might be shared by several
files, which makes the per-file thing doesn't work. For a fs using
trees, it's even more hard to distinguish a file's metadata

quoted

quoted

Also, is this a privileged operation?  If not, then that might be a
problem - could it be used by unprivileged users to work out which
files have been opened recently or something like that?

it's harmless even a unprivileged user uses it. I don't think
unprivileged user can decode the data returned from the ioctl.

um.

Well, by doing a before-and-after thing I can use this ioctl to work
out what metadata blocks are used when someone reads
/my/super/secret-directory/foo.  Then I can write a program which sits
there waiting until someone else reads /my/super/secret-directory/foo. 
Then I can use that information to start WWIII or something.

I dunno, strange things happen.  Unless there's a good *need* to make
this available to unprivileged users then we should not do so.

ok, looks interesting, I'll update the patch to limit unprivileged
users.

Thanks,
Shaohua

On Thu, 20 Jan 2011 10:48:33 +0800 Shaohua Li [off-list ref] wrote:

On Thu, 2011-01-20 at 10:42 +0800, Andrew Morton wrote:

quoted

On Thu, 20 Jan 2011 10:30:47 +0800 Shaohua Li [off-list ref] wrote:

quoted

quoted

I don't know if this is worth addressing.  Perhaps require that the
filp refers to the root of the fs?

I didn't see why this is needed, but I can limit the fip to the root of
the fs.

I don't think it matters much either.  The only problem I can see is if
we were to later try to extend the ioctl into a per-file thing.

since we return page range, a metadata page might be shared by several
files, which makes the per-file thing doesn't work. For a fs using
trees, it's even more hard to distinguish a file's metadata

hm, why.  A query for "which blocks need to be read to access this
file" may return blocks which are shared with other files, but it's
still useful info.  Because it will represent vastly less data (and
hence IO) than the current fs-wide thing.

Now I actually look at it, I cannot find any documentation for the ioctl!  

It seems to return a single offset/length tuple which refers to the
btrfs metadata "file", with the intent that this tuple later be fed
into a btrfs-specific readahead ioctl.

I can see how this might be used with say fatfs or ext3 where all
metadata resides within the blockdev address_space.  But how is a
filesytem which keeps its metadata in multiple address_spaces supposed
to use this interface?

So.  Please fully document the proposed userspace APIs!  This should be
the first thing we look at.  Then we can take a look at how applicable
that is to other-than-btrfs filesystems.