Re: [PATCH v4 38/68] vfs, cachefiles: Mark a backing file in use with an inode flag

[Amir Goldstein]

On Sat, Dec 25, 2021 at 1:32 AM David Howells <dhowells@xxxxxxxxxx> wrote:
>
> Use an inode flag, S_KERNEL_FILE, to mark that a backing file is in use by
> the kernel to prevent cachefiles or other kernel services from interfering
> with that file.
>
> Alter rmdir to reject attempts to remove a directory marked with this flag.
> This is used by cachefiles to prevent cachefilesd from removing them.
>
> Using S_SWAPFILE instead isn't really viable as that has other effects in
> the I/O paths.
>
> Changes
> =======
> ver #3:
>  - Check for the object pointer being NULL in the tracepoints rather than
>    the caller.
>
> Signed-off-by: David Howells <dhowells@xxxxxxxxxx>
> cc: linux-cachefs@xxxxxxxxxx
> Link: https://lore.kernel.org/r/163819630256.215744.4815885535039369574.stgit@xxxxxxxxxxxxxxxxxxxxxx/ # v1
> Link: https://lore.kernel.org/r/163906931596.143852.8642051223094013028.stgit@xxxxxxxxxxxxxxxxxxxxxx/ # v2
> Link: https://lore.kernel.org/r/163967141000.1823006.12920680657559677789.stgit@xxxxxxxxxxxxxxxxxxxxxx/ # v3
> ---
>
>  fs/cachefiles/Makefile            |    1 +
>  fs/cachefiles/namei.c             |   43 +++++++++++++++++++++++++++++++++++++
>  fs/namei.c                        |    3 ++-
>  include/linux/fs.h                |    1 +
>  include/trace/events/cachefiles.h |   42 ++++++++++++++++++++++++++++++++++++
>  5 files changed, 89 insertions(+), 1 deletion(-)
>  create mode 100644 fs/cachefiles/namei.c
>
> diff --git a/fs/cachefiles/Makefile b/fs/cachefiles/Makefile
> index 463e3d608b75..e0b092ca077f 100644
> --- a/fs/cachefiles/Makefile
> +++ b/fs/cachefiles/Makefile
> @@ -7,6 +7,7 @@ cachefiles-y := \
>         cache.o \
>         daemon.o \
>         main.o \
> +       namei.o \
>         security.o
>
>  cachefiles-$(CONFIG_CACHEFILES_ERROR_INJECTION) += error_inject.o
> diff --git a/fs/cachefiles/namei.c b/fs/cachefiles/namei.c
> new file mode 100644
> index 000000000000..913f83f1c900
> --- /dev/null
> +++ b/fs/cachefiles/namei.c
> @@ -0,0 +1,43 @@
> +// SPDX-License-Identifier: GPL-2.0-or-later
> +/* CacheFiles path walking and related routines
> + *
> + * Copyright (C) 2021 Red Hat, Inc. All Rights Reserved.
> + * Written by David Howells (dhowells@xxxxxxxxxx)
> + */
> +
> +#include <linux/fs.h>
> +#include "internal.h"
> +
> +/*
> + * Mark the backing file as being a cache file if it's not already in use.  The
> + * mark tells the culling request command that it's not allowed to cull the
> + * file or directory.  The caller must hold the inode lock.
> + */
> +static bool __cachefiles_mark_inode_in_use(struct cachefiles_object *object,
> +                                          struct dentry *dentry)
> +{
> +       struct inode *inode = d_backing_inode(dentry);
> +       bool can_use = false;
> +
> +       if (!(inode->i_flags & S_KERNEL_FILE)) {
> +               inode->i_flags |= S_KERNEL_FILE;
> +               trace_cachefiles_mark_active(object, inode);
> +               can_use = true;
> +       } else {
> +               pr_notice("cachefiles: Inode already in use: %pd\n", dentry);
> +       }
> +
> +       return can_use;
> +}
> +
> +/*
> + * Unmark a backing inode.  The caller must hold the inode lock.
> + */
> +static void __cachefiles_unmark_inode_in_use(struct cachefiles_object *object,
> +                                            struct dentry *dentry)
> +{
> +       struct inode *inode = d_backing_inode(dentry);
> +
> +       inode->i_flags &= ~S_KERNEL_FILE;
> +       trace_cachefiles_mark_inactive(object, inode);
> +}
> diff --git a/fs/namei.c b/fs/namei.c
> index 1f9d2187c765..d81f04f8d818 100644
> --- a/fs/namei.c
> +++ b/fs/namei.c
> @@ -3958,7 +3958,8 @@ int vfs_rmdir(struct user_namespace *mnt_userns, struct inode *dir,
>         inode_lock(dentry->d_inode);
>
>         error = -EBUSY;
> -       if (is_local_mountpoint(dentry))
> +       if (is_local_mountpoint(dentry) ||
> +           (dentry->d_inode->i_flags & S_KERNEL_FILE))

Better as this check to the many other checks in may_delete()

>                 goto out;
>
>         error = security_inode_rmdir(dir, dentry);
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index 2c0b8e77d9ab..bcf1ca430139 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -2249,6 +2249,7 @@ struct super_operations {
>  #define S_ENCRYPTED    (1 << 14) /* Encrypted file (using fs/crypto/) */
>  #define S_CASEFOLD     (1 << 15) /* Casefolded file */
>  #define S_VERITY       (1 << 16) /* Verity file (using fs/verity/) */
> +#define S_KERNEL_FILE  (1 << 17) /* File is in use by the kernel (eg. fs/cachefiles) */
>

Trying to brand this flag as a generic "in use by kernel" is misleading.
Modules other than cachefiles cannot set/clear this flag, because then
cachefiles won't know that it is allowed to set/clear the flag.

So I think it would be better to call it for what it really is - an inode flag
that is controlled by cachefiles.
Also, the name KERNEL_FILE for a directory is a bit confusing IMO.
Perhaps S_CACHEFILES?

Thanks,
Amir.