Multiply ::name into "mut_name" and "name" which is "const char*".
PDE's name must not be mutated on live PDE, hint modules they should not
do it.
Many other members must not be mutated live as well, but start with
obvious one.
Signed-off-by: Alexey Dobriyan <[email protected]>
---
fs/proc/generic.c | 2 +-
fs/proc/internal.h | 5 ++++-
fs/proc/proc_net.c | 2 +-
3 files changed, 6 insertions(+), 3 deletions(-)
--- a/fs/proc/generic.c
+++ b/fs/proc/generic.c
@@ -436,7 +436,7 @@ static struct proc_dir_entry *__proc_create(struct proc_dir_entry **parent,
}
}
- memcpy(ent->name, fn, qstr.len + 1);
+ memcpy(ent->mut_name, fn, qstr.len + 1);
ent->namelen = qstr.len;
ent->mode = mode;
ent->nlink = nlink;
--- a/fs/proc/internal.h
+++ b/fs/proc/internal.h
@@ -59,7 +59,10 @@ struct proc_dir_entry {
struct proc_dir_entry *parent;
struct rb_root subdir;
struct rb_node subdir_node;
- char *name;
+ union {
+ const char *name;
+ char *mut_name;
+ };
umode_t mode;
u8 flags;
u8 namelen;
--- a/fs/proc/proc_net.c
+++ b/fs/proc/proc_net.c
@@ -368,7 +368,7 @@ static __net_init int proc_net_ns_init(struct net *net)
netd->namelen = 3;
netd->parent = &proc_root;
netd->name = netd->inline_name;
- memcpy(netd->name, "net", 4);
+ memcpy(netd->mut_name, "net", 4);
uid = make_kuid(net->user_ns, 0);
if (!uid_valid(uid))
On Wed, 29 Nov 2023 12:46:57 +0300 Alexey Dobriyan <[email protected]> wrote:
> Multiply ::name into "mut_name" and "name" which is "const char*".
>
> PDE's name must not be mutated on live PDE, hint modules they should not
> do it.
Do any modules do this? If so, we just broke them. If not, why bother
with this change?
On Wed, Nov 29, 2023 at 02:07:05PM -0800, Andrew Morton wrote:
> On Wed, 29 Nov 2023 12:46:57 +0300 Alexey Dobriyan <[email protected]> wrote:
>
> > Multiply ::name into "mut_name" and "name" which is "const char*".
> >
> > PDE's name must not be mutated on live PDE, hint modules they should not
> > do it.
>
> Do any modules do this?
x86_64 allmodconfig is OK, so in-kernel modules are OK.
> If so, we just broke them.
pff...
> If not, why bother with this change?
I don't know, out of love for humanity and increasing OOT modules
code quality. I don't think I've seen changing pde->name, but then
it is OOT modules we are talking about.
C doesn't have a notion for write once members, oh well.