Re: [PATCH 3/6] [RFC] Allow atomic_open() on positive dentry

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 





On 8/16/23 17:25, Miklos Szeredi wrote:
On Wed, 16 Aug 2023 at 16:34, Bernd Schubert <bschubert@xxxxxxx> wrote:

From: Miklos Szeredi <miklos@xxxxxxxxxx>

atomic_open() will do an open-by-name or create-and-open
depending on the flags.

If file was created, then the old positive dentry is obviously
stale, so it will be invalidated and a new one will be allocated.

If not created, then check whether it's the same inode (same as in
->d_revalidate()) and if not, invalidate & allocate new dentry.

Changes (v7 global series) from Miklos initial patch (by Bernd):
- LOOKUP_ATOMIC_REVALIDATE was added and is set for revalidate
   calls into the file system when revalidate by atomic open is
   supported - this is to avoid that ->d_revalidate() would skip
   revalidate and set DCACHE_ATOMIC_OPEN, although vfs
   does not supported it in the given code path (for example
   when LOOKUP_RCU is set)).
- Support atomic-open-revalidate in lookup_fast() - allow atomic
   open for positive dentries without O_CREAT being set.

Changes (v8 global series)
- Introduce enum for d_revalidate return values
- LOOKUP_ATOMIC_REVALIDATE is removed again
- DCACHE_ATOMIC_OPEN flag is replaced by D_REVALIDATE_ATOMIC
   return value

Co-developed-by: Bernd Schubert <bschubert@xxxxxxx>
Signed-off-by: Miklos Szeredi <miklos@xxxxxxxxxx>
Signed-off-by: Bernd Schubert <bschubert@xxxxxxx>
Cc: Christian Brauner <brauner@xxxxxxxxxx>
Cc: Al Viro <viro@xxxxxxxxxxxxxxxxxx>
Cc: Dharmendra Singh <dsingh@xxxxxxx>
Cc: linux-fsdevel@xxxxxxxxxxxxxxx
---
  fs/namei.c            | 25 +++++++++++++++++++------
  include/linux/namei.h |  6 ++++++
  2 files changed, 25 insertions(+), 6 deletions(-)

diff --git a/fs/namei.c b/fs/namei.c
index e4fe0879ae55..8381ec7645f5 100644
--- a/fs/namei.c
+++ b/fs/namei.c
@@ -858,7 +858,7 @@ static inline int d_revalidate(struct dentry *dentry, unsigned int flags)
         if (unlikely(dentry->d_flags & DCACHE_OP_REVALIDATE))
                 return dentry->d_op->d_revalidate(dentry, flags);
         else
-               return 1;
+               return D_REVALIDATE_VALID;
  }

  /**
@@ -1611,10 +1611,11 @@ struct dentry *lookup_one_qstr_excl(const struct qstr *name,
  }
  EXPORT_SYMBOL(lookup_one_qstr_excl);

-static struct dentry *lookup_fast(struct nameidata *nd)
+static struct dentry *lookup_fast(struct nameidata *nd, int *atomic_revalidate)

bool?

  {
         struct dentry *dentry, *parent = nd->path.dentry;
         int status = 1;
+       *atomic_revalidate = 0;

         /*
          * Rename seqlock is not required here because in the off chance
@@ -1656,6 +1657,10 @@ static struct dentry *lookup_fast(struct nameidata *nd)
                 dput(dentry);
                 return ERR_PTR(status);
         }
+
+       if (status == D_REVALIDATE_ATOMIC)
+               *atomic_revalidate = 1;
+
         return dentry;
  }

@@ -1981,6 +1986,7 @@ static const char *handle_dots(struct nameidata *nd, int type)
  static const char *walk_component(struct nameidata *nd, int flags)
  {
         struct dentry *dentry;
+       int atomic_revalidate;
         /*
          * "." and ".." are special - ".." especially so because it has
          * to be able to know about the current root directory and
@@ -1991,7 +1997,7 @@ static const char *walk_component(struct nameidata *nd, int flags)
                         put_link(nd);
                 return handle_dots(nd, nd->last_type);
         }
-       dentry = lookup_fast(nd);
+       dentry = lookup_fast(nd, &atomic_revalidate);
         if (IS_ERR(dentry))
                 return ERR_CAST(dentry);
         if (unlikely(!dentry)) {
@@ -1999,6 +2005,9 @@ static const char *walk_component(struct nameidata *nd, int flags)
                 if (IS_ERR(dentry))
                         return ERR_CAST(dentry);
         }
+
+       WARN_ON(atomic_revalidate);
+
         if (!(flags & WALK_MORE) && nd->depth)
                 put_link(nd);
         return step_into(nd, flags, dentry);
@@ -3430,7 +3439,7 @@ static struct dentry *lookup_open(struct nameidata *nd, struct file *file,
                 dput(dentry);
                 dentry = NULL;
         }
-       if (dentry->d_inode) {
+       if (dentry->d_inode && error != D_REVALIDATE_ATOMIC) {
                 /* Cached positive dentry: will open in f_op->open */
                 return dentry;
         }
@@ -3523,15 +3532,19 @@ static const char *open_last_lookups(struct nameidata *nd,
         }

         if (!(open_flag & O_CREAT)) {
+               int atomic_revalidate;
                 if (nd->last.name[nd->last.len])
                         nd->flags |= LOOKUP_FOLLOW | LOOKUP_DIRECTORY;
                 /* we _can_ be in RCU mode here */
-               dentry = lookup_fast(nd);
+               dentry = lookup_fast(nd, &atomic_revalidate);
                 if (IS_ERR(dentry))
                         return ERR_CAST(dentry);
+               if (dentry && unlikely(atomic_revalidate)) {

Need to assert !LOOKUP_RCU

Are you sure? There is the BUG_ON(nd->flags & LOOKUP_RCU) directly after - should be enough?


+                       dput(dentry);
+                       dentry = NULL;
+               }

Feels a shame to throw away the dentry.  May be worth adding a helper
for the plain atomic open, most of the complexity of lookup_open() is
because of O_CREAT, so this should be much simplified.

Thanks, I'm going to look into it.


                 if (likely(dentry))
                         goto finish_lookup;
-

Adding/removing empty lines is just a distraction, so it shouldn't be
done unless it serves a real purpose.

Ah sorry, accidentally. I'm going to travel the next two days, going to update it on Monday (or at best over the weekend).

Thanks,
Bernd



[Index of Archives]     [Linux Ext4 Filesystem]     [Union Filesystem]     [Filesystem Testing]     [Ceph Users]     [Ecryptfs]     [NTFS 3]     [AutoFS]     [Kernel Newbies]     [Share Photos]     [Security]     [Netfilter]     [Bugtraq]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux Cachefs]     [Reiser Filesystem]     [Linux RAID]     [NTFS 3]     [Samba]     [Device Mapper]     [CEPH Development]

  Powered by Linux