RE: kernel crash in mknod

[Roberto Sassu <roberto.sassu@huawei.com>]

> From: Christian Brauner [mailto:brauner@kernel.org]
> Sent: Monday, March 25, 2024 5:06 PM
> On Sun, Mar 24, 2024 at 04:50:24PM +0000, Roberto Sassu wrote:
> > > From: Al Viro [mailto:viro@ftp.linux.org.uk] On Behalf Of Al Viro
> > > Sent: Sunday, March 24, 2024 6:47 AM
> > > On Sun, Mar 24, 2024 at 12:00:15AM -0500, Steve French wrote:
> > > > Anyone else seeing this kernel crash in do_mknodat (I see it with a
> > > > simple "mkfifo" on smb3 mount).  I started seeing this in 6.9-rc (did
> > > > not see it in 6.8).   I did not see it with the 3/12/23 mainline
> > > > (early in the 6.9-rc merge Window) but I do see it in the 3/22 build
> > > > so it looks like the regression was introduced by:
> > >
> > > 	FWIW, successful ->mknod() is allowed to return 0 and unhash
> > > dentry, rather than bothering with lookups.  So commit in question
> > > is bogus - lack of error does *NOT* mean that you have struct inode
> > > existing, let alone attached to dentry.  That kind of behaviour
> > > used to be common for network filesystems more than just for ->mknod(),
> > > the theory being "if somebody wants to look at it, they can bloody
> > > well pay the cost of lookup after dcache miss".
> > >
> > > Said that, the language in D/f/vfs.rst is vague as hell and is very easy
> > > to misread in direction of "you must instantiate".
> > >
> > > Thankfully, there's no counterpart with mkdir - *there* it's not just
> > > possible, it's inevitable in some cases for e.g. nfs.
> > >
> > > What the hell is that hook doing in non-S_IFREG cases, anyway?  Move it
> > > up and be done with it...
> >
> > Hi Al
> >
> > thanks for the patch. Indeed, it was like that before, when instead of
> > an LSM hook there was an IMA call.
> 
> Could you please start adding lore links into your commit messages for
> all messages that are sent to a mailing list? It really makes tracking
> down the original thread a lot easier.

Sure, will do next time.

> > However, I thought, since we were promoting it as an LSM hook,
> > we should be as generic possible, and support more usages than
> > what was needed for IMA.
> 
> I'm a bit confused now why this is taking a dentry. Nothing in IMA or
> EVM cares about the dentry for these hooks so it really should have take
> an inode in the first place?

Uhm, you are right. Does that mean that instead of what Al proposed,
we can change the parameter of security_path_post_mknod() from
dentry to inode?

> And one minor other question I just realized. Why are some of the new
> hooks called security_path_post_mknod() when they aren't actually taking
> a path in contrast to say
> security_path_{chown,chmod,mknod,chroot,truncate}() that do.

I would agree to any change that makes this more consistent, as long as
IMA has access to the new inode.

Roberto