| Message ID | 20200722090758.3221812-1-chirantan@chromium.org (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [RESEND,PATCHv4,1/2] uapi: fuse: Add FUSE_SECURITY_CTX | expand |
On Wed, Jul 22, 2020 at 6:09 PM Chirantan Ekbote <chirantan@chromium.org> wrote: > > Add the FUSE_SECURITY_CTX flag for the `flags` field of the > fuse_init_out struct. When this flag is set the kernel will append the > security context for a newly created inode to the request (create, > mkdir, mknod, and symlink). The server is responsible for ensuring that > the inode appears atomically with the requested security context. > > For example, if the server is backed by a "real" linux file system then > it can write the security context value to > /proc/thread-self/attr/fscreate before making the syscall to create the > inode. > Friendly ping. Will this (and the next patch in the series) be merged into 5.9? Chirantan Chirantan
On 8/13/2020 10:20 PM, Chirantan Ekbote wrote: > On Wed, Jul 22, 2020 at 6:09 PM Chirantan Ekbote <chirantan@chromium.org> wrote: >> Add the FUSE_SECURITY_CTX flag for the `flags` field of the >> fuse_init_out struct. When this flag is set the kernel will append the >> security context for a newly created inode to the request (create, >> mkdir, mknod, and symlink). The server is responsible for ensuring that >> the inode appears atomically with the requested security context. >> >> For example, if the server is backed by a "real" linux file system then >> it can write the security context value to >> /proc/thread-self/attr/fscreate before making the syscall to create the >> inode. >> > Friendly ping. Will this (and the next patch in the series) be merged into 5.9? This really needed to go to the LSM List <linux-security-module@vger.kernel.org>. > > Chirantan > > > Chirantan
On Wed, Jul 22, 2020 at 06:07:57PM +0900, Chirantan Ekbote wrote: > Add the FUSE_SECURITY_CTX flag for the `flags` field of the > fuse_init_out struct. When this flag is set the kernel will append the > security context for a newly created inode to the request (create, > mkdir, mknod, and symlink). The server is responsible for ensuring that > the inode appears atomically with the requested security context. > > For example, if the server is backed by a "real" linux file system then > it can write the security context value to > /proc/thread-self/attr/fscreate before making the syscall to create the > inode. > > Signed-off-by: Chirantan Ekbote <chirantan@chromium.org> Hi Chirantan, I am wondering what's the status of this work now. Looks like it was not merged. We also need the capability to set selinux security xattrs on newly created files in virtiofs. Will you be interested in reviving this work and send patches again and copy the selinux as well as linux security module list (linux-security-module@vger.kernel.org) as suggested by casey. How are you managing in the meantime. Carrying patches in your own kernel? Thanks Vivek > --- > Changes in v4: > * Added signoff to commit message. > > include/uapi/linux/fuse.h | 8 +++++++- > 1 file changed, 7 insertions(+), 1 deletion(-) > > diff --git a/include/uapi/linux/fuse.h b/include/uapi/linux/fuse.h > index 373cada898159..e2099b45fd44b 100644 > --- a/include/uapi/linux/fuse.h > +++ b/include/uapi/linux/fuse.h > @@ -172,6 +172,10 @@ > * - add FUSE_WRITE_KILL_PRIV flag > * - add FUSE_SETUPMAPPING and FUSE_REMOVEMAPPING > * - add map_alignment to fuse_init_out, add FUSE_MAP_ALIGNMENT flag > + * > + * 7.32 > + * - add FUSE_SECURITY_CTX flag for fuse_init_out > + * - add security context to create, mkdir, symlink, and mknod requests > */ > > #ifndef _LINUX_FUSE_H > @@ -207,7 +211,7 @@ > #define FUSE_KERNEL_VERSION 7 > > /** Minor version number of this interface */ > -#define FUSE_KERNEL_MINOR_VERSION 31 > +#define FUSE_KERNEL_MINOR_VERSION 32 > > /** The node ID of the root inode */ > #define FUSE_ROOT_ID 1 > @@ -314,6 +318,7 @@ struct fuse_file_lock { > * FUSE_NO_OPENDIR_SUPPORT: kernel supports zero-message opendir > * FUSE_EXPLICIT_INVAL_DATA: only invalidate cached pages on explicit request > * FUSE_MAP_ALIGNMENT: map_alignment field is valid > + * FUSE_SECURITY_CTX: add security context to create, mkdir, symlink, and mknod > */ > #define FUSE_ASYNC_READ (1 << 0) > #define FUSE_POSIX_LOCKS (1 << 1) > @@ -342,6 +347,7 @@ struct fuse_file_lock { > #define FUSE_NO_OPENDIR_SUPPORT (1 << 24) > #define FUSE_EXPLICIT_INVAL_DATA (1 << 25) > #define FUSE_MAP_ALIGNMENT (1 << 26) > +#define FUSE_SECURITY_CTX (1 << 27) > > /** > * CUSE INIT request/reply flags > -- > 2.27.0.383.g050319c2ae-goog >
Hi Vivek, On Tue, Jun 15, 2021 at 6:28 AM Vivek Goyal <vgoyal@redhat.com> wrote: > > On Wed, Jul 22, 2020 at 06:07:57PM +0900, Chirantan Ekbote wrote: > > Add the FUSE_SECURITY_CTX flag for the `flags` field of the > > fuse_init_out struct. When this flag is set the kernel will append the > > security context for a newly created inode to the request (create, > > mkdir, mknod, and symlink). The server is responsible for ensuring that > > the inode appears atomically with the requested security context. > > > > For example, if the server is backed by a "real" linux file system then > > it can write the security context value to > > /proc/thread-self/attr/fscreate before making the syscall to create the > > inode. > > > > Signed-off-by: Chirantan Ekbote <chirantan@chromium.org> > > Hi Chirantan, > > I am wondering what's the status of this work now. Looks like it > was not merged. > > We also need the capability to set selinux security xattrs on newly > created files in virtiofs. > > Will you be interested in reviving this work and send patches again > and copy the selinux as well as linux security module list > (linux-security-module@vger.kernel.org) as suggested by casey. > Not really. We have our own local solution for this (see below) so if you or someone else wants to pick it up, please go ahead. > How are you managing in the meantime. Carrying patches in your own > kernel? > Kind of. This patch series changes the protocol and the feature bit we were using was claimed by FUSE_SUBMOUNTS instead so carrying it locally is not really viable long term. Instead we're carrying a patch similar to the original RFC patch that doesn't change the protocol [1]. Chirantan [1]: https://chromium-review.googlesource.com/c/chromiumos/third_party/kernel/+/2581172
On Tue, Jun 15, 2021 at 06:35:21PM +0900, Chirantan Ekbote wrote: > Hi Vivek, > > On Tue, Jun 15, 2021 at 6:28 AM Vivek Goyal <vgoyal@redhat.com> wrote: > > > > On Wed, Jul 22, 2020 at 06:07:57PM +0900, Chirantan Ekbote wrote: > > > Add the FUSE_SECURITY_CTX flag for the `flags` field of the > > > fuse_init_out struct. When this flag is set the kernel will append the > > > security context for a newly created inode to the request (create, > > > mkdir, mknod, and symlink). The server is responsible for ensuring that > > > the inode appears atomically with the requested security context. > > > > > > For example, if the server is backed by a "real" linux file system then > > > it can write the security context value to > > > /proc/thread-self/attr/fscreate before making the syscall to create the > > > inode. > > > > > > Signed-off-by: Chirantan Ekbote <chirantan@chromium.org> > > > > Hi Chirantan, > > > > I am wondering what's the status of this work now. Looks like it > > was not merged. > > > > We also need the capability to set selinux security xattrs on newly > > created files in virtiofs. > > > > Will you be interested in reviving this work and send patches again > > and copy the selinux as well as linux security module list > > (linux-security-module@vger.kernel.org) as suggested by casey. > > > > Not really. We have our own local solution for this (see below) so if > you or someone else wants to pick it up, please go ahead. > Ok. > > How are you managing in the meantime. Carrying patches in your own > > kernel? > > > > Kind of. This patch series changes the protocol and the feature bit we > were using was claimed by FUSE_SUBMOUNTS instead so carrying it > locally is not really viable long term. Instead we're carrying a > patch similar to the original RFC patch that doesn't change the > protocol [1]. Ok, got it. So you went ahead for simpler solution of setting security xattr after creating file hence making it non-atomic. But changelog suggests that it works for your use case as you always do a restorecon on reboot. But I guess upstream will need a solution where file creation and security xattr setting can be atomic. Thanks. If time permits, I might look into the patches you had posted. Thanks Vivek
diff --git a/include/uapi/linux/fuse.h b/include/uapi/linux/fuse.h index 373cada898159..e2099b45fd44b 100644 --- a/include/uapi/linux/fuse.h +++ b/include/uapi/linux/fuse.h @@ -172,6 +172,10 @@ * - add FUSE_WRITE_KILL_PRIV flag * - add FUSE_SETUPMAPPING and FUSE_REMOVEMAPPING * - add map_alignment to fuse_init_out, add FUSE_MAP_ALIGNMENT flag + * + * 7.32 + * - add FUSE_SECURITY_CTX flag for fuse_init_out + * - add security context to create, mkdir, symlink, and mknod requests */ #ifndef _LINUX_FUSE_H @@ -207,7 +211,7 @@ #define FUSE_KERNEL_VERSION 7 /** Minor version number of this interface */ -#define FUSE_KERNEL_MINOR_VERSION 31 +#define FUSE_KERNEL_MINOR_VERSION 32 /** The node ID of the root inode */ #define FUSE_ROOT_ID 1 @@ -314,6 +318,7 @@ struct fuse_file_lock { * FUSE_NO_OPENDIR_SUPPORT: kernel supports zero-message opendir * FUSE_EXPLICIT_INVAL_DATA: only invalidate cached pages on explicit request * FUSE_MAP_ALIGNMENT: map_alignment field is valid + * FUSE_SECURITY_CTX: add security context to create, mkdir, symlink, and mknod */ #define FUSE_ASYNC_READ (1 << 0) #define FUSE_POSIX_LOCKS (1 << 1) @@ -342,6 +347,7 @@ struct fuse_file_lock { #define FUSE_NO_OPENDIR_SUPPORT (1 << 24) #define FUSE_EXPLICIT_INVAL_DATA (1 << 25) #define FUSE_MAP_ALIGNMENT (1 << 26) +#define FUSE_SECURITY_CTX (1 << 27) /** * CUSE INIT request/reply flags
Add the FUSE_SECURITY_CTX flag for the `flags` field of the fuse_init_out struct. When this flag is set the kernel will append the security context for a newly created inode to the request (create, mkdir, mknod, and symlink). The server is responsible for ensuring that the inode appears atomically with the requested security context. For example, if the server is backed by a "real" linux file system then it can write the security context value to /proc/thread-self/attr/fscreate before making the syscall to create the inode. Signed-off-by: Chirantan Ekbote <chirantan@chromium.org> --- Changes in v4: * Added signoff to commit message. include/uapi/linux/fuse.h | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)