We've aligned setgid behavior over multiple kernel releases. The details
can be found in the following two merge messages:
cf619f891971 ("Merge tag 'fs.ovl.setgid.v6.2')
426b4ca2d6a5 ("Merge tag 'fs.setgid.v6.0')
Consistent setgid stripping behavior is now encapsulated in the
setattr_should_drop_sgid() helper which is used by all filesystems that
strip setgid bits outside of vfs proper. Switch nfs to rely on this
helper as well. Without this patch the setgid stripping tests in
xfstests will fail.
Signed-off-by: Christian Brauner (Microsoft) <[email protected]>
---
Changes in v2:
- Christoph Hellwig <[email protected]>:
* Export setattr_should_sgid() so it actually can be used by filesystems
- Link to v1: https://lore.kernel.org/r/[email protected]
---
fs/attr.c | 1 +
fs/internal.h | 2 --
fs/nfs/inode.c | 4 +---
include/linux/fs.h | 2 ++
4 files changed, 4 insertions(+), 5 deletions(-)
diff --git a/fs/attr.c b/fs/attr.c
index aca9ff7aed33..d60dc1edb526 100644
--- a/fs/attr.c
+++ b/fs/attr.c
@@ -47,6 +47,7 @@ int setattr_should_drop_sgid(struct mnt_idmap *idmap,
return ATTR_KILL_SGID;
return 0;
}
+EXPORT_SYMBOL(setattr_should_drop_sgid);
/**
* setattr_should_drop_suidgid - determine whether the set{g,u}id bit needs to
diff --git a/fs/internal.h b/fs/internal.h
index dc4eb91a577a..ab36ed8fa41c 100644
--- a/fs/internal.h
+++ b/fs/internal.h
@@ -259,8 +259,6 @@ ssize_t __kernel_write_iter(struct file *file, struct iov_iter *from, loff_t *po
/*
* fs/attr.c
*/
-int setattr_should_drop_sgid(struct mnt_idmap *idmap,
- const struct inode *inode);
struct mnt_idmap *alloc_mnt_idmap(struct user_namespace *mnt_userns);
struct mnt_idmap *mnt_idmap_get(struct mnt_idmap *idmap);
void mnt_idmap_put(struct mnt_idmap *idmap);
diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
index 222a28320e1c..97a76706fd54 100644
--- a/fs/nfs/inode.c
+++ b/fs/nfs/inode.c
@@ -717,9 +717,7 @@ void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr,
if ((attr->ia_valid & ATTR_KILL_SUID) != 0 &&
inode->i_mode & S_ISUID)
inode->i_mode &= ~S_ISUID;
- if ((attr->ia_valid & ATTR_KILL_SGID) != 0 &&
- (inode->i_mode & (S_ISGID | S_IXGRP)) ==
- (S_ISGID | S_IXGRP))
+ if (setattr_should_drop_sgid(&nop_mnt_idmap, inode))
inode->i_mode &= ~S_ISGID;
if ((attr->ia_valid & ATTR_MODE) != 0) {
int mode = attr->ia_mode & S_IALLUGO;
diff --git a/include/linux/fs.h b/include/linux/fs.h
index c85916e9f7db..af95b64fc810 100644
--- a/include/linux/fs.h
+++ b/include/linux/fs.h
@@ -2675,6 +2675,8 @@ extern struct inode *new_inode(struct super_block *sb);
extern void free_inode_nonrcu(struct inode *inode);
extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *);
extern int file_remove_privs(struct file *);
+int setattr_should_drop_sgid(struct mnt_idmap *idmap,
+ const struct inode *inode);
/*
* This must be used for allocating filesystems specific inodes to set
---
base-commit: eeac8ede17557680855031c6f305ece2378af326
change-id: 20230313-fs-nfs-setgid-659410a10b25
On Tue, Mar 14, 2023 at 12:37:03PM +0000, Matthew Wilcox wrote:
> On Tue, Mar 14, 2023 at 12:51:10PM +0100, Christian Brauner wrote:
> > +++ b/include/linux/fs.h
> > @@ -2675,6 +2675,8 @@ extern struct inode *new_inode(struct super_block *sb);
> > extern void free_inode_nonrcu(struct inode *inode);
> > extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *);
> > extern int file_remove_privs(struct file *);
> > +int setattr_should_drop_sgid(struct mnt_idmap *idmap,
> > + const struct inode *inode);
>
> Is there a reason that setattr_should_drop_sgid() doesn't take a
> const struct mnt_idmap pointer?
Yeah, unfortunately a bunch of helpers that we're passing it down to
from setattr_should_drop_sgid() currently take a non-const pointer to
mnt_idmap. I have a ToDo that to go through all helpers and make them
const wherever we can.
On Tue, Mar 14, 2023 at 12:51:10PM +0100, Christian Brauner wrote:
> +++ b/include/linux/fs.h
> @@ -2675,6 +2675,8 @@ extern struct inode *new_inode(struct super_block *sb);
> extern void free_inode_nonrcu(struct inode *inode);
> extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *);
> extern int file_remove_privs(struct file *);
> +int setattr_should_drop_sgid(struct mnt_idmap *idmap,
> + const struct inode *inode);
Is there a reason that setattr_should_drop_sgid() doesn't take a
const struct mnt_idmap pointer?
On Tue, Mar 14, 2023 at 12:51:10PM +0100, Christian Brauner wrote:
> We've aligned setgid behavior over multiple kernel releases. The details
> can be found in the following two merge messages:
> cf619f891971 ("Merge tag 'fs.ovl.setgid.v6.2')
> 426b4ca2d6a5 ("Merge tag 'fs.setgid.v6.0')
> Consistent setgid stripping behavior is now encapsulated in the
> setattr_should_drop_sgid() helper which is used by all filesystems that
> strip setgid bits outside of vfs proper. Switch nfs to rely on this
> helper as well. Without this patch the setgid stripping tests in
> xfstests will fail.
>
> Signed-off-by: Christian Brauner (Microsoft) <[email protected]>
> ---
> Changes in v2:
> - Christoph Hellwig <[email protected]>:
> * Export setattr_should_sgid() so it actually can be used by filesystems
> - Link to v1: https://lore.kernel.org/r/[email protected]
> ---
If someone has a few cycles to give this a review it would be
appreciated. I'm happy to carry this patch or the NFS tree can.
I'm not fussed.
Looks good:
Reviewed-by: Christoph Hellwig <[email protected]>
On Tue, 14 Mar 2023 12:51:10 +0100, Christian Brauner wrote:
> We've aligned setgid behavior over multiple kernel releases. The details
> can be found in the following two merge messages:
> cf619f891971 ("Merge tag 'fs.ovl.setgid.v6.2')
> 426b4ca2d6a5 ("Merge tag 'fs.setgid.v6.0')
> Consistent setgid stripping behavior is now encapsulated in the
> setattr_should_drop_sgid() helper which is used by all filesystems that
> strip setgid bits outside of vfs proper. Switch nfs to rely on this
> helper as well. Without this patch the setgid stripping tests in
> xfstests will fail.
>
> [...]
So I've picked this up now,
tree: git://git.kernel.org/pub/scm/linux/kernel/git/vfs/idmapping.git
branch: fs.misc
[1/1] nfs: use vfs setgid helper
commit: 4f704d9a8352f5c0a8fcdb6213b934630342bd44
Thanks!
Christian
Hi Christian,
On Tue, Mar 14, 2023 at 7:51 AM Christian Brauner <[email protected]> wrote:
>
> We've aligned setgid behavior over multiple kernel releases. The details
> can be found in the following two merge messages:
> cf619f891971 ("Merge tag 'fs.ovl.setgid.v6.2')
> 426b4ca2d6a5 ("Merge tag 'fs.setgid.v6.0')
> Consistent setgid stripping behavior is now encapsulated in the
> setattr_should_drop_sgid() helper which is used by all filesystems that
> strip setgid bits outside of vfs proper. Switch nfs to rely on this
> helper as well. Without this patch the setgid stripping tests in
> xfstests will fail.
>
> Signed-off-by: Christian Brauner (Microsoft) <[email protected]>
> ---
> Changes in v2:
> - Christoph Hellwig <[email protected]>:
> * Export setattr_should_sgid() so it actually can be used by filesystems
> - Link to v1: https://lore.kernel.org/r/[email protected]
> ---
> fs/attr.c | 1 +
> fs/internal.h | 2 --
> fs/nfs/inode.c | 4 +---
> include/linux/fs.h | 2 ++
> 4 files changed, 4 insertions(+), 5 deletions(-)
>
> diff --git a/fs/attr.c b/fs/attr.c
> index aca9ff7aed33..d60dc1edb526 100644
> --- a/fs/attr.c
> +++ b/fs/attr.c
> @@ -47,6 +47,7 @@ int setattr_should_drop_sgid(struct mnt_idmap *idmap,
> return ATTR_KILL_SGID;
> return 0;
> }
> +EXPORT_SYMBOL(setattr_should_drop_sgid);
>
> /**
> * setattr_should_drop_suidgid - determine whether the set{g,u}id bit needs to
> diff --git a/fs/internal.h b/fs/internal.h
> index dc4eb91a577a..ab36ed8fa41c 100644
> --- a/fs/internal.h
> +++ b/fs/internal.h
> @@ -259,8 +259,6 @@ ssize_t __kernel_write_iter(struct file *file, struct iov_iter *from, loff_t *po
> /*
> * fs/attr.c
> */
> -int setattr_should_drop_sgid(struct mnt_idmap *idmap,
> - const struct inode *inode);
> struct mnt_idmap *alloc_mnt_idmap(struct user_namespace *mnt_userns);
> struct mnt_idmap *mnt_idmap_get(struct mnt_idmap *idmap);
> void mnt_idmap_put(struct mnt_idmap *idmap);
> diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
> index 222a28320e1c..97a76706fd54 100644
> --- a/fs/nfs/inode.c
> +++ b/fs/nfs/inode.c
> @@ -717,9 +717,7 @@ void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr,
> if ((attr->ia_valid & ATTR_KILL_SUID) != 0 &&
> inode->i_mode & S_ISUID)
> inode->i_mode &= ~S_ISUID;
> - if ((attr->ia_valid & ATTR_KILL_SGID) != 0 &&
> - (inode->i_mode & (S_ISGID | S_IXGRP)) ==
> - (S_ISGID | S_IXGRP))
> + if (setattr_should_drop_sgid(&nop_mnt_idmap, inode))
> inode->i_mode &= ~S_ISGID;
> if ((attr->ia_valid & ATTR_MODE) != 0) {
> int mode = attr->ia_mode & S_IALLUGO;
Will this be going through your tree (due to the VFS leve changes)?
If so, you can add:
Acked-by: Anna Schumaker <[email protected]>
for the NFS bits.
Thanks,
Anna
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index c85916e9f7db..af95b64fc810 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -2675,6 +2675,8 @@ extern struct inode *new_inode(struct super_block *sb);
> extern void free_inode_nonrcu(struct inode *inode);
> extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *);
> extern int file_remove_privs(struct file *);
> +int setattr_should_drop_sgid(struct mnt_idmap *idmap,
> + const struct inode *inode);
>
> /*
> * This must be used for allocating filesystems specific inodes to set
>
> ---
> base-commit: eeac8ede17557680855031c6f305ece2378af326
> change-id: 20230313-fs-nfs-setgid-659410a10b25
>
On Tue, 2023-03-14 at 12:51 +0100, Christian Brauner wrote:
> We've aligned setgid behavior over multiple kernel releases. The details
> can be found in the following two merge messages:
> cf619f891971 ("Merge tag 'fs.ovl.setgid.v6.2')
> 426b4ca2d6a5 ("Merge tag 'fs.setgid.v6.0')
> Consistent setgid stripping behavior is now encapsulated in the
> setattr_should_drop_sgid() helper which is used by all filesystems that
> strip setgid bits outside of vfs proper. Switch nfs to rely on this
> helper as well. Without this patch the setgid stripping tests in
> xfstests will fail.
>
> Signed-off-by: Christian Brauner (Microsoft) <[email protected]>
> ---
> Changes in v2:
> - Christoph Hellwig <[email protected]>:
> * Export setattr_should_sgid() so it actually can be used by filesystems
> - Link to v1: https://lore.kernel.org/r/[email protected]
> ---
> fs/attr.c | 1 +
> fs/internal.h | 2 --
> fs/nfs/inode.c | 4 +---
> include/linux/fs.h | 2 ++
> 4 files changed, 4 insertions(+), 5 deletions(-)
>
> diff --git a/fs/attr.c b/fs/attr.c
> index aca9ff7aed33..d60dc1edb526 100644
> --- a/fs/attr.c
> +++ b/fs/attr.c
> @@ -47,6 +47,7 @@ int setattr_should_drop_sgid(struct mnt_idmap *idmap,
> return ATTR_KILL_SGID;
> return 0;
> }
> +EXPORT_SYMBOL(setattr_should_drop_sgid);
>
> /**
> * setattr_should_drop_suidgid - determine whether the set{g,u}id bit needs to
> diff --git a/fs/internal.h b/fs/internal.h
> index dc4eb91a577a..ab36ed8fa41c 100644
> --- a/fs/internal.h
> +++ b/fs/internal.h
> @@ -259,8 +259,6 @@ ssize_t __kernel_write_iter(struct file *file, struct iov_iter *from, loff_t *po
> /*
> * fs/attr.c
> */
> -int setattr_should_drop_sgid(struct mnt_idmap *idmap,
> - const struct inode *inode);
> struct mnt_idmap *alloc_mnt_idmap(struct user_namespace *mnt_userns);
> struct mnt_idmap *mnt_idmap_get(struct mnt_idmap *idmap);
> void mnt_idmap_put(struct mnt_idmap *idmap);
> diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
> index 222a28320e1c..97a76706fd54 100644
> --- a/fs/nfs/inode.c
> +++ b/fs/nfs/inode.c
> @@ -717,9 +717,7 @@ void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr,
> if ((attr->ia_valid & ATTR_KILL_SUID) != 0 &&
> inode->i_mode & S_ISUID)
> inode->i_mode &= ~S_ISUID;
> - if ((attr->ia_valid & ATTR_KILL_SGID) != 0 &&
> - (inode->i_mode & (S_ISGID | S_IXGRP)) ==
> - (S_ISGID | S_IXGRP))
> + if (setattr_should_drop_sgid(&nop_mnt_idmap, inode))
> inode->i_mode &= ~S_ISGID;
> if ((attr->ia_valid & ATTR_MODE) != 0) {
> int mode = attr->ia_mode & S_IALLUGO;
> diff --git a/include/linux/fs.h b/include/linux/fs.h
> index c85916e9f7db..af95b64fc810 100644
> --- a/include/linux/fs.h
> +++ b/include/linux/fs.h
> @@ -2675,6 +2675,8 @@ extern struct inode *new_inode(struct super_block *sb);
> extern void free_inode_nonrcu(struct inode *inode);
> extern int setattr_should_drop_suidgid(struct mnt_idmap *, struct inode *);
> extern int file_remove_privs(struct file *);
> +int setattr_should_drop_sgid(struct mnt_idmap *idmap,
> + const struct inode *inode);
>
> /*
> * This must be used for allocating filesystems specific inodes to set
>
> ---
> base-commit: eeac8ede17557680855031c6f305ece2378af326
> change-id: 20230313-fs-nfs-setgid-659410a10b25
>
Reviewed-by: Jeff Layton <[email protected]>
On Tue, Apr 11, 2023 at 12:54:25PM -0400, Anna Schumaker wrote:
> Hi Christian,
>
> On Tue, Mar 14, 2023 at 7:51 AM Christian Brauner <[email protected]> wrote:
> >
> > We've aligned setgid behavior over multiple kernel releases. The details
> > can be found in the following two merge messages:
> > cf619f891971 ("Merge tag 'fs.ovl.setgid.v6.2')
> > 426b4ca2d6a5 ("Merge tag 'fs.setgid.v6.0')
> > Consistent setgid stripping behavior is now encapsulated in the
> > setattr_should_drop_sgid() helper which is used by all filesystems that
> > strip setgid bits outside of vfs proper. Switch nfs to rely on this
> > helper as well. Without this patch the setgid stripping tests in
> > xfstests will fail.
> >
> > Signed-off-by: Christian Brauner (Microsoft) <[email protected]>
> > ---
> > Changes in v2:
> > - Christoph Hellwig <[email protected]>:
> > * Export setattr_should_sgid() so it actually can be used by filesystems
> > - Link to v1: https://lore.kernel.org/r/[email protected]
> > ---
> > fs/attr.c | 1 +
> > fs/internal.h | 2 --
> > fs/nfs/inode.c | 4 +---
> > include/linux/fs.h | 2 ++
> > 4 files changed, 4 insertions(+), 5 deletions(-)
> >
> > diff --git a/fs/attr.c b/fs/attr.c
> > index aca9ff7aed33..d60dc1edb526 100644
> > --- a/fs/attr.c
> > +++ b/fs/attr.c
> > @@ -47,6 +47,7 @@ int setattr_should_drop_sgid(struct mnt_idmap *idmap,
> > return ATTR_KILL_SGID;
> > return 0;
> > }
> > +EXPORT_SYMBOL(setattr_should_drop_sgid);
> >
> > /**
> > * setattr_should_drop_suidgid - determine whether the set{g,u}id bit needs to
> > diff --git a/fs/internal.h b/fs/internal.h
> > index dc4eb91a577a..ab36ed8fa41c 100644
> > --- a/fs/internal.h
> > +++ b/fs/internal.h
> > @@ -259,8 +259,6 @@ ssize_t __kernel_write_iter(struct file *file, struct iov_iter *from, loff_t *po
> > /*
> > * fs/attr.c
> > */
> > -int setattr_should_drop_sgid(struct mnt_idmap *idmap,
> > - const struct inode *inode);
> > struct mnt_idmap *alloc_mnt_idmap(struct user_namespace *mnt_userns);
> > struct mnt_idmap *mnt_idmap_get(struct mnt_idmap *idmap);
> > void mnt_idmap_put(struct mnt_idmap *idmap);
> > diff --git a/fs/nfs/inode.c b/fs/nfs/inode.c
> > index 222a28320e1c..97a76706fd54 100644
> > --- a/fs/nfs/inode.c
> > +++ b/fs/nfs/inode.c
> > @@ -717,9 +717,7 @@ void nfs_setattr_update_inode(struct inode *inode, struct iattr *attr,
> > if ((attr->ia_valid & ATTR_KILL_SUID) != 0 &&
> > inode->i_mode & S_ISUID)
> > inode->i_mode &= ~S_ISUID;
> > - if ((attr->ia_valid & ATTR_KILL_SGID) != 0 &&
> > - (inode->i_mode & (S_ISGID | S_IXGRP)) ==
> > - (S_ISGID | S_IXGRP))
> > + if (setattr_should_drop_sgid(&nop_mnt_idmap, inode))
> > inode->i_mode &= ~S_ISGID;
> > if ((attr->ia_valid & ATTR_MODE) != 0) {
> > int mode = attr->ia_mode & S_IALLUGO;
>
> Will this be going through your tree (due to the VFS leve changes)?
> If so, you can add:
>
> Acked-by: Anna Schumaker <[email protected]>
Yes, I have it applied to fs.misc. There should be a corresponding
"applied message" I've sent last week. Thanks for taking a look!
Christian