public inbox for linux-unionfs@vger.kernel.org
 help / color / mirror / Atom feed
From: Stefan Berger <stefanb@linux.ibm.com>
To: Amir Goldstein <amir73il@gmail.com>
Cc: Christian Brauner <brauner@kernel.org>,
	Mimi Zohar <zohar@linux.ibm.com>,
	Jeff Layton <jlayton@kernel.org>,
	Casey Schaufler <casey@schaufler-ca.com>,
	syzbot <syzbot+a67fc5321ffb4b311c98@syzkaller.appspotmail.com>,
	linux-fsdevel@vger.kernel.org, linux-integrity@vger.kernel.org,
	linux-kernel@vger.kernel.org, linux-unionfs@vger.kernel.org,
	miklos@szeredi.hu, syzkaller-bugs@googlegroups.com
Subject: Re: [syzbot] [integrity] [overlayfs] general protection fault in d_path
Date: Fri, 29 Sep 2023 08:39:48 -0400	[thread overview]
Message-ID: <99294acf-7275-8f4d-a129-d5df208b7b2a@linux.ibm.com> (raw)
In-Reply-To: <CAOQ4uxie6xT5mmCcCwYtnEvra37eSeFftXfxaTULfdJnk1VcXQ@mail.gmail.com>


On 9/29/23 00:25, Amir Goldstein wrote:
> On Fri, Sep 29, 2023 at 3:02 AM Stefan Berger <stefanb@linux.ibm.com> wrote:
>>
>> On 9/21/23 07:48, Christian Brauner wrote:
>>> Imho, this is all very wild but I'm not judging.
>>>
>>> Two solutions imho:
>>> (1) teach stacking filesystems like overlayfs and ecryptfs to use
>>>       vfs_getattr_nosec() in their ->getattr() implementation when they
>>>       are themselves called via vfs_getattr_nosec(). This will fix this by
>>>       not triggering another LSM hook.
>>
>> You can avoid all this churn.
>> Just use the existing query_flags arg.
>> Nothing outside the AT_STATX_SYNC_TYPE query_flags is
>> passed into filesystems from userspace.
>>
>> Mast out AT_STATX_SYNC_TYPE in vfs_getattr()
>> And allow kernel internal request_flags in vfs_getattr_nosec()
Hm, I thought that vfs_getattr_nosec needs to pass AT_GETATTR_NOSEC into 
->getattr().
>>
>> The AT_ flag namespace is already a challenge, but mixing user
>> flags and kernel-only flags in vfs interfaces has been done before.
>>
>> ...


That's what I wanted to avoid since now all filesystems' getattr() may 
have the AT_GETATTR_NOSEC mixed into the query_flags.

Anyway, here's what I currently have:

diff --git a/fs/ecryptfs/inode.c b/fs/ecryptfs/inode.c
index 992d9c7e64ae..f7b5b1843dcc 100644
--- a/fs/ecryptfs/inode.c
+++ b/fs/ecryptfs/inode.c
@@ -998,16 +998,28 @@ static int ecryptfs_getattr_link(struct mnt_idmap 
*idmap,
         return rc;
  }

+static int ecryptfs_do_getattr(bool nosec, const struct path *path,
+                              struct kstat *stat, u32 request_mask,
+                              unsigned int flags)
+{
+       if (nosec)
+               return vfs_getattr_nosec(path, stat, request_mask, flags);
+       return vfs_getattr(path, stat, request_mask, flags);
+}
+
  static int ecryptfs_getattr(struct mnt_idmap *idmap,
                             const struct path *path, struct kstat *stat,
                             u32 request_mask, unsigned int flags)
  {
         struct dentry *dentry = path->dentry;
         struct kstat lower_stat;
+       bool nosec = flags & AT_GETATTR_NOSEC;
         int rc;

-       rc = vfs_getattr(ecryptfs_dentry_to_lower_path(dentry), &lower_stat,
-                        request_mask, flags);
+       flags &= ~AT_INTERNAL_MASK;
+
+       rc = ecryptfs_do_getattr(nosec, 
ecryptfs_dentry_to_lower_path(dentry),
+                                &lower_stat, request_mask, flags);
         if (!rc) {
                 fsstack_copy_attr_all(d_inode(dentry),
ecryptfs_inode_to_lower(d_inode(dentry)));
diff --git a/fs/overlayfs/inode.c b/fs/overlayfs/inode.c
index 83ef66644c21..ec4ceb5b4ebf 100644
--- a/fs/overlayfs/inode.c
+++ b/fs/overlayfs/inode.c
@@ -166,12 +166,15 @@ int ovl_getattr(struct mnt_idmap *idmap, const 
struct path *path,
         int fsid = 0;
         int err;
         bool metacopy_blocks = false;
+       bool nosec = flags & AT_GETATTR_NOSEC;
+
+       flags &= ~AT_INTERNAL_MASK;

         metacopy_blocks = ovl_is_metacopy_dentry(dentry);

         type = ovl_path_real(dentry, &realpath);
         old_cred = ovl_override_creds(dentry->d_sb);
-       err = vfs_getattr(&realpath, stat, request_mask, flags);
+       err = ovl_do_getattr(nosec, &realpath, stat, request_mask, flags);
         if (err)
                 goto out;

@@ -196,8 +199,8 @@ int ovl_getattr(struct mnt_idmap *idmap, const 
struct path *path,
                                         (!is_dir ? STATX_NLINK : 0);

                         ovl_path_lower(dentry, &realpath);
-                       err = vfs_getattr(&realpath, &lowerstat,
-                                         lowermask, flags);
+                       err = ovl_do_getattr(nosec, &realpath, &lowerstat,
+                                            lowermask, flags);
                         if (err)
                                 goto out;

@@ -249,8 +252,9 @@ int ovl_getattr(struct mnt_idmap *idmap, const 
struct path *path,

                         ovl_path_lowerdata(dentry, &realpath);
                         if (realpath.dentry) {
-                               err = vfs_getattr(&realpath, &lowerdatastat,
-                                                 lowermask, flags);
+                               err = ovl_do_getattr(nosec, &realpath,
+ &lowerdatastat, lowermask,
+                                                    flags);
                                 if (err)
                                         goto out;
                         } else {
diff --git a/fs/overlayfs/overlayfs.h b/fs/overlayfs/overlayfs.h
index 9817b2dcb132..cbee3ff3bab7 100644
--- a/fs/overlayfs/overlayfs.h
+++ b/fs/overlayfs/overlayfs.h
@@ -397,6 +397,15 @@ static inline bool ovl_open_flags_need_copy_up(int 
flags)
         return ((OPEN_FMODE(flags) & FMODE_WRITE) || (flags & O_TRUNC));
  }

+static inline int ovl_do_getattr(bool nosec, const struct path *path,
+                                struct kstat *stat, u32 request_mask,
+                                unsigned int flags)
+{
+       if (nosec)
+               return vfs_getattr_nosec(path, stat, request_mask, flags);
+       return vfs_getattr(path, stat, request_mask, flags);
+}
+
  /* util.c */
  int ovl_want_write(struct dentry *dentry);
  void ovl_drop_write(struct dentry *dentry);
diff --git a/fs/stat.c b/fs/stat.c
index d43a5cc1bfa4..3250e427e1aa 100644
--- a/fs/stat.c
+++ b/fs/stat.c
@@ -133,7 +133,8 @@ int vfs_getattr_nosec(const struct path *path, 
struct kstat *stat,
         idmap = mnt_idmap(path->mnt);
         if (inode->i_op->getattr)
                 return inode->i_op->getattr(idmap, path, stat,
-                                           request_mask, query_flags);
+                                           request_mask,
+                                           query_flags | AT_GETATTR_NOSEC);

         generic_fillattr(idmap, request_mask, inode, stat);
         return 0;
diff --git a/include/linux/fs.h b/include/linux/fs.h
index b528f063e8ff..9069d6a301f0 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -2027,6 +2027,12 @@ struct super_operations {
         void (*shutdown)(struct super_block *sb);
  };

+/*
+ * Internal query flags. See fcntl.h AT_xxx flags for the rest.
+ */
+#define AT_GETATTR_NOSEC               0x80000000
+#define AT_INTERNAL_MASK               0x80000000
+
  /*
   * Inode flags - they have no relation to superblock flags now
   */




  reply	other threads:[~2023-09-29 12:40 UTC|newest]

Thread overview: 35+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2023-07-05 12:00 [syzbot] [overlayfs?] general protection fault in d_path syzbot
2023-07-05 13:05 ` Christian Brauner
2023-07-05 13:39   ` Amir Goldstein
2023-07-05 13:41   ` Jeff Layton
2023-07-05 13:54     ` Amir Goldstein
2023-07-05 14:10       ` Jeff Layton
2023-09-12 22:10 ` syzbot
2023-09-13  7:09   ` Amir Goldstein
2023-09-13  9:06     ` [syzbot] [integrity] [overlayfs] " syzbot
2023-09-14 17:54       ` Mimi Zohar
2023-09-18  0:04 ` syzbot
2023-09-20 17:01   ` Stefan Berger
2023-09-20 20:37     ` Stefan Berger
2023-09-20 21:16       ` Jeff Layton
2023-09-20 22:09         ` Stefan Berger
2023-09-21  0:10           ` Stefan Berger
2023-09-21  0:52             ` Casey Schaufler
2023-09-21 10:32               ` Jeff Layton
2023-09-21 11:24                 ` Mimi Zohar
2023-09-21 11:48                   ` Christian Brauner
2023-09-21 14:29                     ` Stefan Berger
2023-09-21 14:52                     ` Mimi Zohar
2023-09-21 15:10                       ` Jeff Layton
2023-09-21 15:18                         ` Amir Goldstein
2023-09-21 15:19                         ` Mimi Zohar
2023-09-21 15:39                           ` Jeff Layton
2023-09-21 16:06                             ` Mimi Zohar
2023-09-21 17:01                               ` Amir Goldstein
2023-09-26 14:40                                 ` Mimi Zohar
2023-09-29  4:30                                   ` Amir Goldstein
2023-09-28 22:03                     ` Stefan Berger
2023-09-29  4:25                       ` Amir Goldstein
2023-09-29 12:39                         ` Stefan Berger [this message]
2023-10-01 14:51                           ` Amir Goldstein
2023-09-21 12:03                 ` Amir Goldstein

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=99294acf-7275-8f4d-a129-d5df208b7b2a@linux.ibm.com \
    --to=stefanb@linux.ibm.com \
    --cc=amir73il@gmail.com \
    --cc=brauner@kernel.org \
    --cc=casey@schaufler-ca.com \
    --cc=jlayton@kernel.org \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-integrity@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-unionfs@vger.kernel.org \
    --cc=miklos@szeredi.hu \
    --cc=syzbot+a67fc5321ffb4b311c98@syzkaller.appspotmail.com \
    --cc=syzkaller-bugs@googlegroups.com \
    --cc=zohar@linux.ibm.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox