samsung_sm8750/android_kernel_samsung_sm8750
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

UPSTREAM: af_unix: Replace garbage collection algorithm.

Browse Source 
commit 4090fa373f0e763c43610853d2774b5979915959 upstream.

If we find a dead SCC during iteration, we call unix_collect_skb()
to splice all skb in the SCC to the global sk_buff_head, hitlist.

After iterating all SCC, we unlock unix_gc_lock and purge the queue.

Bug: 404256079
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Acked-by: Paolo Abeni <pabeni@redhat.com>
Link: https://lore.kernel.org/r/20240325202425.60930-15-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Lee Jones <lee@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
(cherry picked from commit de7921631f)
Signed-off-by: Lee Jones <joneslee@google.com>
Change-Id: Ia3698af8bf0d00775e4a18fea6b137c9fac70173
This commit is contained in:
Kuniyuki Iwashima
2025-05-21 14:45:28 +00:00
committed by Lee Jones
parent 67a3a58da1
commit 11d208f893
5 changed files with 94 additions and 315 deletions
Download Patch File Download Diff File Expand all files Collapse all files

61
android/abi_gki_aarch64.stg
View File

@@ -100498,12 +100498,6 @@ member {
type_id: 0x246497bd type_id: 0x246497bd
offset: 128 offset: 128
} }
member {
id: 0x805723f7
name: "gc_flags"
type_id: 0x33756485
offset: 8256
}
member { member {
id: 0x8fd944b0 id: 0x8fd944b0
name: "gc_interval" name: "gc_interval"
@@ -112771,12 +112765,6 @@ member {
type_id: 0x6d7f5ff6 type_id: 0x6d7f5ff6
offset: 16320 offset: 16320
} }
member {
id: 0x79c0ce48
name: "inflight"
type_id: 0x33756485
offset: 8128
}
member { member {
id: 0x79f74293 id: 0x79f74293
name: "inflight" name: "inflight"
@@ -123480,12 +123468,6 @@ member {
name: "link" name: "link"
type_id: 0xd3c80119 type_id: 0xd3c80119
} }
member {
id: 0x556d5b03
name: "link"
type_id: 0xd3c80119
offset: 8000
}
member { member {
id: 0x556d5d6f id: 0x556d5d6f
name: "link" name: "link"
@@ -125010,6 +124992,12 @@ member {
name: "lock" name: "lock"
type_id: 0xf4933b90 type_id: 0xf4933b90
} }
member {
id: 0x2d1fe06a
name: "lock"
type_id: 0xf313e71a
offset: 8000
}
member { member {
id: 0x2d1fe165 id: 0x2d1fe165
name: "lock" name: "lock"
@@ -125202,12 +125190,6 @@ member {
type_id: 0xf313e71a type_id: 0xf313e71a
offset: 320 offset: 320
} }
member {
id: 0x2d1feb84
name: "lock"
type_id: 0xf313e71a
offset: 8192
}
member { member {
id: 0x2d1feb9a id: 0x2d1feb9a
name: "lock" name: "lock"
@@ -148618,10 +148600,10 @@ member {
offset: 896 offset: 896
} }
member { member {
id: 0x337b4b40 id: 0x337b4570
name: "oob_skb" name: "oob_skb"
type_id: 0x054f691a type_id: 0x054f691a
offset: 9600 offset: 9088
} }
member { member {
id: 0x97510416 id: 0x97510416
@@ -154605,16 +154587,16 @@ member {
offset: 304 offset: 304
} }
member { member {
id: 0x6a6786bf id: 0x6a6789c8
name: "peer_wake" name: "peer_wake"
type_id: 0x347ff86e type_id: 0x347ff86e
offset: 9216 offset: 8704
} }
member { member {
id: 0x337f3f92 id: 0x337f3641
name: "peer_wq" name: "peer_wq"
type_id: 0x5eee2044 type_id: 0x5eee2044
offset: 8704 offset: 8192
} }
member { member {
id: 0xc29a8766 id: 0xc29a8766
@@ -180668,10 +180650,10 @@ member {
type_id: 0xc9082b19 type_id: 0xc9082b19
} }
member { member {
id: 0xac8121db id: 0xac812894
name: "scm_stat" name: "scm_stat"
type_id: 0xfa80d3fa type_id: 0xfa80d3fa
offset: 9536 offset: 9024
} }
member { member {
id: 0xe94f5bce id: 0xe94f5bce
@@ -277497,7 +277479,7 @@ struct_union {
kind: STRUCT kind: STRUCT
name: "unix_sock" name: "unix_sock"
definition { definition {
bytesize: 1216 bytesize: 1152
member_id: 0x82ce9da8 member_id: 0x82ce9da8
member_id: 0x24257f11 member_id: 0x24257f11
member_id: 0x77353e9c member_id: 0x77353e9c
@@ -277506,14 +277488,11 @@ struct_union {
member_id: 0xb794a532 member_id: 0xb794a532
member_id: 0x5eddc09d member_id: 0x5eddc09d
member_id: 0xe2a2a0b6 member_id: 0xe2a2a0b6
member_id: 0x556d5b03 member_id: 0x2d1fe06a
member_id: 0x79c0ce48 member_id: 0x337f3641
member_id: 0x2d1feb84 member_id: 0x6a6789c8
member_id: 0x805723f7 member_id: 0xac812894
member_id: 0x337f3f92 member_id: 0x337b4570
member_id: 0x6a6786bf
member_id: 0xac8121db
member_id: 0x337b4b40
} }
} }
struct_union { struct_union {

10
android/abi_gki_aarch64.stg.allowed_breaks
View File

@@ -192,3 +192,13 @@ type 'struct unix_vertex' changed
member 'unsigned long lowlink' was removed member 'unsigned long lowlink' was removed
member 'unsigned long scc_index' was added member 'unsigned long scc_index' was added
type 'struct unix_sock' changed
byte size changed from 1216 to 1152
member 'struct list_head link' was removed
member 'unsigned long inflight' was removed
member 'spinlock_t lock' changed
offset changed by -192
member 'unsigned long gc_flags' was removed
4 members ('struct socket_wq peer_wq' .. 'struct sk_buff* oob_skb') changed
offset changed by -512

8
include/net/af_unix.h
View File

@@ -19,9 +19,6 @@ static inline struct unix_sock *unix_get_socket(struct file *filp)
extern spinlock_t unix_gc_lock; extern spinlock_t unix_gc_lock;
extern unsigned int unix_tot_inflight; 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_add_edges(struct scm_fp_list *fpl, struct unix_sock *receiver); void unix_add_edges(struct scm_fp_list *fpl, struct unix_sock *receiver);
void unix_del_edges(struct scm_fp_list *fpl); void unix_del_edges(struct scm_fp_list *fpl);
void unix_update_edges(struct unix_sock *receiver); void unix_update_edges(struct unix_sock *receiver);
@@ -85,12 +82,7 @@ struct unix_sock {
struct sock *peer; struct sock *peer;
struct unix_vertex *vertex; struct unix_vertex *vertex;
struct sock *listener; struct sock *listener;
struct list_head link;
unsigned long inflight;
spinlock_t lock; spinlock_t lock;
unsigned long gc_flags;
#define UNIX_GC_CANDIDATE 0
#define UNIX_GC_MAYBE_CYCLE 1
struct socket_wq peer_wq; struct socket_wq peer_wq;
wait_queue_entry_t peer_wake; wait_queue_entry_t peer_wake;
struct scm_stat scm_stat; struct scm_stat scm_stat;

12
net/unix/af_unix.c
View File

@@ -979,12 +979,10 @@ static struct sock *unix_create1(struct net *net, struct socket *sock, int kern,
sk->sk_destruct = unix_sock_destructor; sk->sk_destruct = unix_sock_destructor;
u = unix_sk(sk); u = unix_sk(sk);
u->listener = NULL; u->listener = NULL;
u->inflight = 0;
u->vertex = NULL; u->vertex = NULL;
u->path.dentry = NULL; u->path.dentry = NULL;
u->path.mnt = NULL; u->path.mnt = NULL;
spin_lock_init(&u->lock); spin_lock_init(&u->lock);
INIT_LIST_HEAD(&u->link);
mutex_init(&u->iolock); /* single task reading lock */ mutex_init(&u->iolock); /* single task reading lock */
mutex_init(&u->bindlock); /* single task binding lock */ mutex_init(&u->bindlock); /* single task binding lock */
init_waitqueue_head(&u->peer_wait); init_waitqueue_head(&u->peer_wait);
@@ -1770,8 +1768,6 @@ static inline bool too_many_unix_fds(struct task_struct *p)
static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb) static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb)
{ {
int i;
if (too_many_unix_fds(current)) if (too_many_unix_fds(current))
return -ETOOMANYREFS; return -ETOOMANYREFS;
@@ -1783,9 +1779,6 @@ static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb)
if (!UNIXCB(skb).fp) if (!UNIXCB(skb).fp)
return -ENOMEM; return -ENOMEM;
for (i = scm->fp->count - 1; i >= 0; i--)
unix_inflight(scm->fp->user, scm->fp->fp[i]);
if (unix_prepare_fpl(UNIXCB(skb).fp)) if (unix_prepare_fpl(UNIXCB(skb).fp))
return -ENOMEM; return -ENOMEM;
@@ -1794,15 +1787,10 @@ static int unix_attach_fds(struct scm_cookie *scm, struct sk_buff *skb)
static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb) static void unix_detach_fds(struct scm_cookie *scm, struct sk_buff *skb)
{ {
int i;
scm->fp = UNIXCB(skb).fp; scm->fp = UNIXCB(skb).fp;
UNIXCB(skb).fp = NULL; UNIXCB(skb).fp = NULL;
unix_destroy_fpl(scm->fp); unix_destroy_fpl(scm->fp);
for (i = scm->fp->count - 1; i >= 0; i--)
unix_notinflight(scm->fp->user, scm->fp->fp[i]);
} }
static void unix_peek_fds(struct scm_cookie *scm, struct sk_buff *skb) static void unix_peek_fds(struct scm_cookie *scm, struct sk_buff *skb)

318
net/unix/garbage.c
View File

@@ -322,6 +322,52 @@ static bool unix_vertex_dead(struct unix_vertex *vertex)
return true; return true;
} }
enum unix_recv_queue_lock_class {
U_RECVQ_LOCK_NORMAL,
U_RECVQ_LOCK_EMBRYO,
};
static void unix_collect_skb(struct list_head *scc, struct sk_buff_head *hitlist)
{
struct unix_vertex *vertex;
list_for_each_entry_reverse(vertex, scc, scc_entry) {
struct sk_buff_head *queue;
struct unix_edge *edge;
struct unix_sock *u;
edge = list_first_entry(&vertex->edges, typeof(*edge), vertex_entry);
u = edge->predecessor;
queue = &u->sk.sk_receive_queue;
spin_lock(&queue->lock);
if (u->sk.sk_state == TCP_LISTEN) {
struct sk_buff *skb;
skb_queue_walk(queue, skb) {
struct sk_buff_head *embryo_queue = &skb->sk->sk_receive_queue;
/* listener -> embryo order, the inversion never happens. */
spin_lock_nested(&embryo_queue->lock, U_RECVQ_LOCK_EMBRYO);
skb_queue_splice_init(embryo_queue, hitlist);
spin_unlock(&embryo_queue->lock);
}
} else {
skb_queue_splice_init(queue, hitlist);
#if IS_ENABLED(CONFIG_AF_UNIX_OOB)
if (u->oob_skb) {
kfree_skb(u->oob_skb);
u->oob_skb = NULL;
}
#endif
}
spin_unlock(&queue->lock);
}
}
static bool unix_scc_cyclic(struct list_head *scc) static bool unix_scc_cyclic(struct list_head *scc)
{ {
struct unix_vertex *vertex; struct unix_vertex *vertex;
@@ -345,7 +391,8 @@ static bool unix_scc_cyclic(struct list_head *scc)
static LIST_HEAD(unix_visited_vertices); static LIST_HEAD(unix_visited_vertices);
static unsigned long unix_vertex_grouped_index = UNIX_VERTEX_INDEX_MARK2; static unsigned long unix_vertex_grouped_index = UNIX_VERTEX_INDEX_MARK2;
static void __unix_walk_scc(struct unix_vertex *vertex, unsigned long *last_index) static void __unix_walk_scc(struct unix_vertex *vertex, unsigned long *last_index,
struct sk_buff_head *hitlist)
{ {
LIST_HEAD(vertex_stack); LIST_HEAD(vertex_stack);
struct unix_edge *edge; struct unix_edge *edge;
@@ -430,7 +477,9 @@ prev_vertex:
scc_dead = unix_vertex_dead(vertex); scc_dead = unix_vertex_dead(vertex);
} }
if (!unix_graph_maybe_cyclic) if (scc_dead)
unix_collect_skb(&scc, hitlist);
else if (!unix_graph_maybe_cyclic)
unix_graph_maybe_cyclic = unix_scc_cyclic(&scc); unix_graph_maybe_cyclic = unix_scc_cyclic(&scc);
list_del(&scc); list_del(&scc);
@@ -441,7 +490,7 @@ prev_vertex:
goto prev_vertex; goto prev_vertex;
} }
static void unix_walk_scc(void) static void unix_walk_scc(struct sk_buff_head *hitlist)
{ {
unsigned long last_index = UNIX_VERTEX_INDEX_START; unsigned long last_index = UNIX_VERTEX_INDEX_START;
@@ -454,7 +503,7 @@ static void unix_walk_scc(void)
struct unix_vertex *vertex; struct unix_vertex *vertex;
vertex = list_first_entry(&unix_unvisited_vertices, typeof(*vertex), entry); vertex = list_first_entry(&unix_unvisited_vertices, typeof(*vertex), entry);
__unix_walk_scc(vertex, &last_index); __unix_walk_scc(vertex, &last_index, hitlist);
} }
list_replace_init(&unix_visited_vertices, &unix_unvisited_vertices); list_replace_init(&unix_visited_vertices, &unix_unvisited_vertices);
@@ -463,7 +512,7 @@ static void unix_walk_scc(void)
unix_graph_grouped = true; unix_graph_grouped = true;
} }
static void unix_walk_scc_fast(void) static void unix_walk_scc_fast(struct sk_buff_head *hitlist)
{ {
while (!list_empty(&unix_unvisited_vertices)) { while (!list_empty(&unix_unvisited_vertices)) {
struct unix_vertex *vertex; struct unix_vertex *vertex;
@@ -480,279 +529,40 @@ static void unix_walk_scc_fast(void)
scc_dead = unix_vertex_dead(vertex); scc_dead = unix_vertex_dead(vertex);
} }
if (scc_dead)
unix_collect_skb(&scc, hitlist);
list_del(&scc); list_del(&scc);
} }
list_replace_init(&unix_visited_vertices, &unix_unvisited_vertices); list_replace_init(&unix_visited_vertices, &unix_unvisited_vertices);
} }
static LIST_HEAD(gc_candidates);
static LIST_HEAD(gc_inflight_list);
/* Keep the number of times in flight count for the file
* descriptor if it is for an AF_UNIX socket.
*/
void unix_inflight(struct user_struct *user, struct file *filp)
{
struct unix_sock *u = unix_get_socket(filp);
spin_lock(&unix_gc_lock);
if (u) {
if (!u->inflight) {
WARN_ON_ONCE(!list_empty(&u->link));
list_add_tail(&u->link, &gc_inflight_list);
} else {
WARN_ON_ONCE(list_empty(&u->link));
}
u->inflight++;
}
spin_unlock(&unix_gc_lock);
}
void unix_notinflight(struct user_struct *user, struct file *filp)
{
struct unix_sock *u = unix_get_socket(filp);
spin_lock(&unix_gc_lock);
if (u) {
WARN_ON_ONCE(!u->inflight);
WARN_ON_ONCE(list_empty(&u->link));
u->inflight--;
if (!u->inflight)
list_del_init(&u->link);
}
spin_unlock(&unix_gc_lock);
}
static void scan_inflight(struct sock *x, void (*func)(struct unix_sock *),
struct sk_buff_head *hitlist)
{
struct sk_buff *skb;
struct sk_buff *next;
spin_lock(&x->sk_receive_queue.lock);
skb_queue_walk_safe(&x->sk_receive_queue, skb, next) {
/* Do we have file descriptors ? */
if (UNIXCB(skb).fp) {
bool hit = false;
/* Process the descriptors of this socket */
int nfd = UNIXCB(skb).fp->count;
struct file **fp = UNIXCB(skb).fp->fp;
while (nfd--) {
/* Get the socket the fd matches if it indeed does so */
struct unix_sock *u = unix_get_socket(*fp++);
/* Ignore non-candidates, they could have been added
* to the queues after starting the garbage collection
*/
if (u && test_bit(UNIX_GC_CANDIDATE, &u->gc_flags)) {
hit = true;
func(u);
}
}
if (hit && hitlist != NULL) {
__skb_unlink(skb, &x->sk_receive_queue);
__skb_queue_tail(hitlist, skb);
}
}
}
spin_unlock(&x->sk_receive_queue.lock);
}
static void scan_children(struct sock *x, void (*func)(struct unix_sock *),
struct sk_buff_head *hitlist)
{
if (x->sk_state != TCP_LISTEN) {
scan_inflight(x, func, hitlist);
} else {
struct sk_buff *skb;
struct sk_buff *next;
struct unix_sock *u;
LIST_HEAD(embryos);
/* For a listening socket collect the queued embryos
* and perform a scan on them as well.
*/
spin_lock(&x->sk_receive_queue.lock);
skb_queue_walk_safe(&x->sk_receive_queue, skb, next) {
u = unix_sk(skb->sk);
/* An embryo cannot be in-flight, so it's safe
* to use the list link.
*/
WARN_ON_ONCE(!list_empty(&u->link));
list_add_tail(&u->link, &embryos);
}
spin_unlock(&x->sk_receive_queue.lock);
while (!list_empty(&embryos)) {
u = list_entry(embryos.next, struct unix_sock, link);
scan_inflight(&u->sk, func, hitlist);
list_del_init(&u->link);
}
}
}
static void dec_inflight(struct unix_sock *usk)
{
usk->inflight--;
}
static void inc_inflight(struct unix_sock *usk)
{
usk->inflight++;
}
static void inc_inflight_move_tail(struct unix_sock *u)
{
u->inflight++;
/* If this still might be part of a cycle, move it to the end
* of the list, so that it's checked even if it was already
* passed over
*/
if (test_bit(UNIX_GC_MAYBE_CYCLE, &u->gc_flags))
list_move_tail(&u->link, &gc_candidates);
}
static bool gc_in_progress; static bool gc_in_progress;
static void __unix_gc(struct work_struct *work) static void __unix_gc(struct work_struct *work)
{ {
struct sk_buff_head hitlist; struct sk_buff_head hitlist;
struct unix_sock *u, *next;
LIST_HEAD(not_cycle_list);
struct list_head cursor;
spin_lock(&unix_gc_lock); spin_lock(&unix_gc_lock);
if (!unix_graph_maybe_cyclic) if (!unix_graph_maybe_cyclic) {
spin_unlock(&unix_gc_lock);
goto skip_gc; goto skip_gc;
}
__skb_queue_head_init(&hitlist);
if (unix_graph_grouped) if (unix_graph_grouped)
unix_walk_scc_fast(); unix_walk_scc_fast(&hitlist);
else else
unix_walk_scc(); unix_walk_scc(&hitlist);
/* First, select candidates for garbage collection. Only
* in-flight sockets are considered, and from those only ones
* which don't have any external reference.
*
* Holding unix_gc_lock will protect these candidates from
* being detached, and hence from gaining an external
* reference. Since there are no possible receivers, all
* buffers currently on the candidates' queues stay there
* during the garbage collection.
*
* We also know that no new candidate can be added onto the
* receive queues. Other, non candidate sockets _can_ be
* added to queue, so we must make sure only to touch
* candidates.
*
* Embryos, though never candidates themselves, affect which
* candidates are reachable by the garbage collector. Before
* being added to a listener's queue, an embryo may already
* receive data carrying SCM_RIGHTS, potentially making the
* passed socket a candidate that is not yet reachable by the
* collector. It becomes reachable once the embryo is
* enqueued. Therefore, we must ensure that no SCM-laden
* embryo appears in a (candidate) listener's queue between
* consecutive scan_children() calls.
*/
list_for_each_entry_safe(u, next, &gc_inflight_list, link) {
struct sock *sk = &u->sk;
long total_refs;
total_refs = file_count(sk->sk_socket->file);
WARN_ON_ONCE(!u->inflight);
WARN_ON_ONCE(total_refs < u->inflight);
if (total_refs == u->inflight) {
list_move_tail(&u->link, &gc_candidates);
__set_bit(UNIX_GC_CANDIDATE, &u->gc_flags);
__set_bit(UNIX_GC_MAYBE_CYCLE, &u->gc_flags);
if (sk->sk_state == TCP_LISTEN) {
unix_state_lock_nested(sk, U_LOCK_GC_LISTENER);
unix_state_unlock(sk);
}
}
}
/* Now remove all internal in-flight reference to children of
* the candidates.
*/
list_for_each_entry(u, &gc_candidates, link)
scan_children(&u->sk, dec_inflight, NULL);
/* Restore the references for children of all candidates,
* which have remaining references. Do this recursively, so
* only those remain, which form cyclic references.
*
* Use a "cursor" link, to make the list traversal safe, even
* though elements might be moved about.
*/
list_add(&cursor, &gc_candidates);
while (cursor.next != &gc_candidates) {
u = list_entry(cursor.next, struct unix_sock, link);
/* Move cursor to after the current position. */
list_move(&cursor, &u->link);
if (u->inflight) {
list_move_tail(&u->link, &not_cycle_list);
__clear_bit(UNIX_GC_MAYBE_CYCLE, &u->gc_flags);
scan_children(&u->sk, inc_inflight_move_tail, NULL);
}
}
list_del(&cursor);
/* Now gc_candidates contains only garbage. Restore original
* inflight counters for these as well, and remove the skbuffs
* which are creating the cycle(s).
*/
skb_queue_head_init(&hitlist);
list_for_each_entry(u, &gc_candidates, link) {
scan_children(&u->sk, inc_inflight, &hitlist);
#if IS_ENABLED(CONFIG_AF_UNIX_OOB)
if (u->oob_skb) {
kfree_skb(u->oob_skb);
u->oob_skb = NULL;
}
#endif
}
/* not_cycle_list contains those sockets which do not make up a
* cycle. Restore these to the inflight list.
*/
while (!list_empty(&not_cycle_list)) {
u = list_entry(not_cycle_list.next, struct unix_sock, link);
__clear_bit(UNIX_GC_CANDIDATE, &u->gc_flags);
list_move_tail(&u->link, &gc_inflight_list);
}
spin_unlock(&unix_gc_lock); spin_unlock(&unix_gc_lock);
/* Here we are. Hitlist is filled. Die. */
__skb_queue_purge(&hitlist); __skb_queue_purge(&hitlist);
spin_lock(&unix_gc_lock);
/* All candidates should have been detached by now. */
WARN_ON_ONCE(!list_empty(&gc_candidates));
skip_gc: skip_gc:
/* Paired with READ_ONCE() in wait_for_unix_gc(). */
WRITE_ONCE(gc_in_progress, false); WRITE_ONCE(gc_in_progress, false);
spin_unlock(&unix_gc_lock);
} }
static DECLARE_WORK(unix_gc_work, __unix_gc); static DECLARE_WORK(unix_gc_work, __unix_gc);