From 7c349ed090318b1c88a3e5dff3b24f732296edce Mon Sep 17 00:00:00 2001 From: Kuniyuki Iwashima Date: Mon, 1 Apr 2024 10:31:24 -0700 Subject: [PATCH 1/2] af_unix: Remove scm_fp_dup() in unix_attach_fds(). When we passed fds, we used to bump each file's refcount twice in scm_fp_copy() and scm_fp_dup() before linking the socket to gc_inflight_list. This is because we incremented the inflight count of the socket and linked it to the list in advance before passing skb to the destination socket. Otherwise, the inflight socket could have been garbage-collected in a small race window between linking the socket to the list and queuing skb: CPU 1 : sendmsg(X) w/ A's fd CPU 2 : close(A) ----- ----- /* Here A's refcount is 1, and inflight count is 0 */ bump A's refcount to 2 in scm_fp_copy() bump A's inflight count to 1 link A to gc_inflight_list decrement A's refcount to 1 /* A's refcount == inflight count, thus A could be GC candidate */ start GC mark A as candidate purge A's receive queue queue skb w/ A's fd to X /* A is queued, but all data has been lost */ After commit 4090fa373f0e ("af_unix: Replace garbage collection algorithm."), we increment the inflight count and link the socket to the global list only when queuing the skb. The race no longer exists, so let's not clone the fd nor bump the count in unix_attach_fds(). Signed-off-by: Kuniyuki Iwashima Link: https://lore.kernel.org/r/20240401173125.92184-2-kuniyu@amazon.com Signed-off-by: Jakub Kicinski --- net/unix/af_unix.c | 9 ++------- 1 file changed, 2 insertions(+), 7 deletions(-) diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c index 533fb682c954..78be8b520cef 100644 --- a/net/unix/af_unix.c +++ b/net/unix/af_unix.c @@ -1794,13 +1794,8 @@ static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb) if (too_many_unix_fds(current)) return -ETOOMANYREFS; - /* Need to duplicate file references for the sake of garbage - * collection. Otherwise a socket in the fps might become a - * candidate for GC while the skb is not yet queued. - */ - UNIXCB(skb).fp = scm_fp_dup(scm->fp); - if (!UNIXCB(skb).fp) - return -ENOMEM; + UNIXCB(skb).fp = scm->fp; + scm->fp = NULL; if (unix_prepare_fpl(UNIXCB(skb).fp)) return -ENOMEM; From 118f457da9ed58a79e24b73c2ef0aa1987241f0e Mon Sep 17 00:00:00 2001 From: Kuniyuki Iwashima Date: Mon, 1 Apr 2024 10:31:25 -0700 Subject: [PATCH 2/2] af_unix: Remove lock dance in unix_peek_fds(). In the previous GC implementation, the shape of the inflight socket graph was not expected to change while GC was in progress. MSG_PEEK was tricky because it could install inflight fd silently and transform the graph. Let's say we peeked a fd, which was a listening socket, and accept()ed some embryo sockets from it. The garbage collection algorithm would have been confused because the set of sockets visited in scan_inflight() would change within the same GC invocation. That's why we placed spin_lock(&unix_gc_lock) and spin_unlock() in unix_peek_fds() with a fat comment. In the new GC implementation, we no longer garbage-collect the socket if it exists in another queue, that is, if it has a bridge to another SCC. Also, accept() will require the lock if it has edges. Thus, we need not do the complicated lock dance. Signed-off-by: Kuniyuki Iwashima Link: https://lore.kernel.org/r/20240401173125.92184-3-kuniyu@amazon.com Signed-off-by: Jakub Kicinski --- include/net/af_unix.h | 1 - net/unix/af_unix.c | 42 ------------------------------------------ net/unix/garbage.c | 2 +- 3 files changed, 1 insertion(+), 44 deletions(-) diff --git a/include/net/af_unix.h b/include/net/af_unix.h index 226a8da2cbe3..7311b77edfc7 100644 --- a/include/net/af_unix.h +++ b/include/net/af_unix.h @@ -17,7 +17,6 @@ static inline struct unix_sock *unix_get_socket(struct file *filp) } #endif -extern spinlock_t unix_gc_lock; extern unsigned int unix_tot_inflight; void unix_add_edges(struct scm_fp_list *fpl, struct unix_sock *receiver); void unix_del_edges(struct scm_fp_list *fpl); diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c index 78be8b520cef..61ecfa9c9c6b 100644 --- a/net/unix/af_unix.c +++ b/net/unix/af_unix.c @@ -1814,48 +1814,6 @@ static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb) static void unix_peek_fds(struct scm_cookie *scm, struct sk_buff *skb) { scm->fp = scm_fp_dup(UNIXCB(skb).fp); - - /* - * Garbage collection of unix sockets starts by selecting a set of - * candidate sockets which have reference only from being in flight - * (total_refs == inflight_refs). This condition is checked once during - * the candidate collection phase, and candidates are marked as such, so - * that non-candidates can later be ignored. While inflight_refs is - * protected by unix_gc_lock, total_refs (file count) is not, hence this - * is an instantaneous decision. - * - * Once a candidate, however, the socket must not be reinstalled into a - * file descriptor while the garbage collection is in progress. - * - * If the above conditions are met, then the directed graph of - * candidates (*) does not change while unix_gc_lock is held. - * - * Any operations that changes the file count through file descriptors - * (dup, close, sendmsg) does not change the graph since candidates are - * not installed in fds. - * - * Dequeing a candidate via recvmsg would install it into an fd, but - * that takes unix_gc_lock to decrement the inflight count, so it's - * serialized with garbage collection. - * - * MSG_PEEK is special in that it does not change the inflight count, - * yet does install the socket into an fd. The following lock/unlock - * pair is to ensure serialization with garbage collection. It must be - * done between incrementing the file count and installing the file into - * an fd. - * - * If garbage collection starts after the barrier provided by the - * lock/unlock, then it will see the elevated refcount and not mark this - * as a candidate. If a garbage collection is already in progress - * before the file count was incremented, then the lock/unlock pair will - * ensure that garbage collection is finished before progressing to - * installing the fd. - * - * (*) A -> B where B is on the queue of A or B is on the queue of C - * which is on the queue of listening socket A. - */ - spin_lock(&unix_gc_lock); - spin_unlock(&unix_gc_lock); } static void unix_destruct_scm(struct sk_buff *skb) diff --git a/net/unix/garbage.c b/net/unix/garbage.c index 89ea71d9297b..12a4ec27e0d4 100644 --- a/net/unix/garbage.c +++ b/net/unix/garbage.c @@ -183,7 +183,7 @@ static void unix_free_vertices(struct scm_fp_list *fpl) } } -DEFINE_SPINLOCK(unix_gc_lock); +static DEFINE_SPINLOCK(unix_gc_lock); unsigned int unix_tot_inflight; void unix_add_edges(struct scm_fp_list *fpl, struct unix_sock *receiver)