| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <20240220174437.47356-1-kuniyu@amazon.com>
Date: Tue, 20 Feb 2024 09:44:37 -0800
From: Kuniyuki Iwashima <kuniyu@...zon.com>
To: <pabeni@...hat.com>
CC: <davem@...emloft.net>, <edumazet@...gle.com>, <kuba@...nel.org>,
<kuni1840@...il.com>, <kuniyu@...zon.com>, <netdev@...r.kernel.org>
Subject: Re: [PATCH v2 net-next 03/14] af_unix: Link struct unix_edge when queuing skb.
From: Paolo Abeni <pabeni@...hat.com>
Date: Tue, 20 Feb 2024 13:11:09 +0100
> On Fri, 2024-02-16 at 13:05 -0800, Kuniyuki Iwashima wrote:
> > Just before queuing skb with inflight fds, we call scm_stat_add(),
> > which is a good place to set up the preallocated struct unix_edge
> > in UNIXCB(skb).fp->edges.
> >
> > Then, we call unix_add_edges() and construct the directed graph
> > as follows:
> >
> > 1. Set the inflight socket's unix_vertex to unix_edge.predecessor
> > 2. Set the receiver's unix_vertex to unix_edge.successor
> > 3. Link unix_edge.entry to the inflight socket's unix_vertex.edges
> > 4. Link inflight socket's unix_vertex.entry to unix_unvisited_vertices.
> >
> > Let's say we pass the fd of AF_UNIX socket A to B and the fd of B
> > to C. The graph looks like this:
> >
> > +-------------------------+
> > | unix_unvisited_vertices | <------------------------.
> > +-------------------------+ |
> > + |
> > | +-------------+ +-------------+ | +-------------+
> > | | unix_sock A | | unix_sock B | | | unix_sock C |
> > | +-------------+ +-------------+ | +-------------+
> > | | unix_vertex | <----. .----> | unix_vertex | <-|--. .----> | unix_vertex |
> > | | +-----------+ | | | +-----------+ | | | | +-----------+
> > `-> | | entry | +------------> | | entry | +-' | | | | entry |
> > | |-----------| | | | |-----------| | | | |-----------|
> > | | edges | <-. | | | | edges | <-. | | | | edges |
> > +-+-----------+ | | | +-+-----------+ | | | +-+-----------+
> > | | | | | |
> > .---------------------' | | .---------------------' | |
> > | | | | | |
> > | +-------------+ | | | +-------------+ | |
> > | | unix_edge | | | | | unix_edge | | |
> > | +-------------+ | | | +-------------+ | |
> > `-> | entry | | | `-> | entry | | |
> > |-------------| | | |-------------| | |
> > | predecessor | +----' | | predecessor | +----' |
> > |-------------| | |-------------| |
> > | successor | +-------' | successor | +-------'
> > +-------------+ +-------------+
> >
> > Henceforth, we denote such a graph as A -> B (-> C).
> >
> > Now, we can express all inflight fd graphs that do not contain
> > embryo sockets. The following two patches will support the
> > particular case.
> >
> > Signed-off-by: Kuniyuki Iwashima <kuniyu@...zon.com>
> > ---
> > include/net/af_unix.h | 2 ++
> > include/net/scm.h | 1 +
> > net/core/scm.c | 2 ++
> > net/unix/af_unix.c | 8 +++++--
> > net/unix/garbage.c | 55 ++++++++++++++++++++++++++++++++++++++++++-
> > 5 files changed, 65 insertions(+), 3 deletions(-)
> >
> > diff --git a/include/net/af_unix.h b/include/net/af_unix.h
> > index cab9dfb666f3..54d62467a70b 100644
> > --- a/include/net/af_unix.h
> > +++ b/include/net/af_unix.h
> > @@ -23,6 +23,8 @@ extern unsigned int unix_tot_inflight;
> > void unix_inflight(struct user_struct *user, struct file *fp);
> > void unix_notinflight(struct user_struct *user, struct file *fp);
> > void unix_init_vertex(struct unix_sock *u);
> > +void unix_add_edges(struct scm_fp_list *fpl, struct unix_sock *receiver);
> > +void unix_del_edges(struct scm_fp_list *fpl);
> > int unix_alloc_edges(struct scm_fp_list *fpl);
> > void unix_free_edges(struct scm_fp_list *fpl);
> > void unix_gc(void);
> > diff --git a/include/net/scm.h b/include/net/scm.h
> > index a1142dee086c..7d807fe466a3 100644
> > --- a/include/net/scm.h
> > +++ b/include/net/scm.h
> > @@ -32,6 +32,7 @@ struct scm_fp_list {
> > short count_unix;
> > short max;
> > #ifdef CONFIG_UNIX
> > + bool inflight;
> > struct unix_edge *edges;
> > #endif
> > struct user_struct *user;
> > diff --git a/net/core/scm.c b/net/core/scm.c
> > index bc75b6927222..cad0c153ac93 100644
> > --- a/net/core/scm.c
> > +++ b/net/core/scm.c
> > @@ -88,6 +88,7 @@ static int scm_fp_copy(struct cmsghdr *cmsg, struct scm_fp_list **fplp)
> > fpl->count = 0;
> > fpl->count_unix = 0;
> > #if IS_ENABLED(CONFIG_UNIX)
> > + fpl->inflight = false;
> > fpl->edges = NULL;
> > #endif
> > fpl->max = SCM_MAX_FD;
> > @@ -381,6 +382,7 @@ struct scm_fp_list *scm_fp_dup(struct scm_fp_list *fpl)
> > get_file(fpl->fp[i]);
> >
> > #if IS_ENABLED(CONFIG_UNIX)
> > + new_fpl->inflight = false;
> > new_fpl->edges = NULL;
> > #endif
> > new_fpl->max = new_fpl->count;
> > diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c
> > index 0391f66546a6..ea7bac18a781 100644
> > --- a/net/unix/af_unix.c
> > +++ b/net/unix/af_unix.c
> > @@ -1956,8 +1956,10 @@ static void scm_stat_add(struct sock *sk, struct sk_buff *skb)
> > struct scm_fp_list *fp = UNIXCB(skb).fp;
> > struct unix_sock *u = unix_sk(sk);
> >
> > - if (unlikely(fp && fp->count))
> > + if (unlikely(fp && fp->count)) {
> > atomic_add(fp->count, &u->scm_stat.nr_fds);
> > + unix_add_edges(fp, u);
> > + }
> > }
> >
> > static void scm_stat_del(struct sock *sk, struct sk_buff *skb)
> > @@ -1965,8 +1967,10 @@ static void scm_stat_del(struct sock *sk, struct sk_buff *skb)
> > struct scm_fp_list *fp = UNIXCB(skb).fp;
> > struct unix_sock *u = unix_sk(sk);
> >
> > - if (unlikely(fp && fp->count))
> > + if (unlikely(fp && fp->count)) {
> > atomic_sub(fp->count, &u->scm_stat.nr_fds);
> > + unix_del_edges(fp);
> > + }
> > }
> >
> > /*
> > diff --git a/net/unix/garbage.c b/net/unix/garbage.c
> > index ec998c7d6b4c..353416f38738 100644
> > --- a/net/unix/garbage.c
> > +++ b/net/unix/garbage.c
> > @@ -109,6 +109,57 @@ void unix_init_vertex(struct unix_sock *u)
> > INIT_LIST_HEAD(&vertex->edges);
> > }
> >
> > +DEFINE_SPINLOCK(unix_gc_lock);
> > +static LIST_HEAD(unix_unvisited_vertices);
> > +
> > +void unix_add_edges(struct scm_fp_list *fpl, struct unix_sock *receiver)
> > +{
> > + int i = 0, j = 0;
> > +
> > + spin_lock(&unix_gc_lock);
> > +
> > + while (i < fpl->count_unix) {
> > + struct unix_sock *inflight = unix_get_socket(fpl->fp[j++]);
> > + struct unix_edge *edge;
> > +
> > + if (!inflight)
> > + continue;
> > +
> > + edge = fpl->edges + i++;
> > + edge->predecessor = &inflight->vertex;
> > + edge->successor = &receiver->vertex;
> > +
> > + if (!edge->predecessor->out_degree++)
> > + list_add_tail(&edge->predecessor->entry, &unix_unvisited_vertices);
> > +
> > + list_add_tail(&edge->entry, &edge->predecessor->edges);
>
> Note that I confusingly replied to the previous revision of this patch,
> but I think the points still stand.
>
> INIT_LIST_HEAD(&edge->entry);
>
> disappeared from the above, but I can't find where it landed?!?
Sorry, I forgot to mention this change in the coverletter.
Initially, I placed INIT_LIST_HEAD() in unix_alloc_edges(), but in v1,
I moved it to unix_add_edges(), and later I noticed it's unnecessary
in the first place because edge->entry is not used as head of a list,
so I removed it completely in v2.
From: Paolo Abeni <pabeni@...hat.com>
Date: Tue, 20 Feb 2024 13:06:18 +0100
> Here 'edge->predecessor->entry' and 'edge->entry' refer to different
> object types right ? edge vs vertices. Perhaps using different field
> names could clarify the code a bit?
Regarding the name of edge->entry, I agree a diffrent name would be
easy to understand. I'll rename it to edge->vertex_entry unless there
is a better name :)
Thanks!
Powered by blists - more mailing lists