This is step 4/4 of a patch series to fix mid_q_entry memory leaks
caused by race conditions in callback execution.
In compound_send_recv(), when wait_for_response() is interrupted by
signals, the code attempts to cancel pending requests by changing
their callbacks to cifs_cancelled_callback. However, there's a race
condition between signal interruption and network response processing
that causes both mid_q_entry and server buffer leaks:
```
User foreground process cifsd
cifs_readdir
open_cached_dir
cifs_send_recv
compound_send_recv
smb2_setup_request
smb2_mid_entry_alloc
smb2_get_mid_entry
smb2_mid_entry_alloc
mempool_alloc // alloc mid
kref_init(&temp->refcount); // refcount = 1
mid[0]->callback = cifs_compound_callback;
mid[1]->callback = cifs_compound_last_callback;
smb_send_rqst
rc = wait_for_response
wait_event_state TASK_KILLABLE
cifs_demultiplex_thread
allocate_buffers
server->bigbuf = cifs_buf_get()
standard_receive3
->find_mid()
smb2_find_mid
__smb2_find_mid
kref_get(&mid->refcount) // +1
cifs_handle_standard
handle_mid
/* bigbuf will also leak */
mid->resp_buf = server->bigbuf
server->bigbuf = NULL;
dequeue_mid
/* in for loop */
mids[0]->callback
cifs_compound_callback
/* Signal interrupts wait: rc = -ERESTARTSYS */
/* if (... || midQ[i]->mid_state == MID_RESPONSE_RECEIVED) *?
midQ[0]->callback = cifs_cancelled_callback;
cancelled_mid[i] = true;
/* The change comes too late */
mid->mid_state = MID_RESPONSE_READY
release_mid // -1
/* cancelled_mid[i] == true causes mid won't be released
in compound_send_recv cleanup */
/* cifs_cancelled_callback won't executed to release mid */
```
The root cause is that there's a race between callback assignment and
execution.
Fix this by introducing per-mid locking:
- Add spinlock_t mid_lock to struct mid_q_entry
- Add mid_execute_callback() for atomic callback execution
- Use mid_lock in cancellation paths to ensure atomicity
This ensures that either the original callback or the cancellation
callback executes atomically, preventing reference count leaks when
requests are interrupted by signals.
Link: https://bugzilla.kernel.org/show_bug.cgi?id=220404
Fixes:
ee258d79159a ("CIFS: Move credit processing to mid callbacks for SMB3")
Signed-off-by: Wang Zhaolong <wangzhaolong@huaweicloud.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
int mid_rc; /* rc for MID_RC */
__le16 command; /* smb command code */
unsigned int optype; /* operation type */
+ spinlock_t mid_lock;
bool wait_cancelled:1; /* Cancelled while waiting for response */
bool deleted_from_q:1; /* Whether Mid has been dequeued frem pending_mid_q */
bool large_buf:1; /* if valid response, is pointer to large buf */
* cifsFileInfo->file_info_lock cifsFileInfo->count cifs_new_fileinfo
* ->invalidHandle initiate_cifs_search
* ->oplock_break_cancelled
+ * mid_q_entry->mid_lock mid_q_entry->callback alloc_mid
+ * smb2_mid_entry_alloc
+ * (Any fields of mid_q_entry that will need protection)
****************************************************************************/
#ifdef DECLARE_GLOBALS_HERE
return ret;
}
+/*
+ * Execute mid callback atomically - ensures callback runs exactly once
+ * and prevents sleeping in atomic context.
+ */
+static inline void mid_execute_callback(struct mid_q_entry *mid)
+{
+ void (*callback)(struct mid_q_entry *mid);
+
+ spin_lock(&mid->mid_lock);
+ callback = mid->callback;
+ mid->callback = NULL; /* Mark as executed, */
+ spin_unlock(&mid->mid_lock);
+
+ if (callback)
+ callback(mid);
+}
+
#define CIFS_REPARSE_SUPPORT(tcon) \
((tcon)->posix_extensions || \
(le32_to_cpu((tcon)->fsAttrInfo.Attributes) & \
temp = mempool_alloc(cifs_mid_poolp, GFP_NOFS);
memset(temp, 0, sizeof(struct mid_q_entry));
kref_init(&temp->refcount);
+ spin_lock_init(&temp->mid_lock);
temp->mid = get_mid(smb_buffer);
temp->pid = current->pid;
temp->command = cpu_to_le16(smb_buffer->Command);
rc = wait_for_response(server, midQ);
if (rc != 0) {
send_cancel(server, &rqst, midQ);
- spin_lock(&server->mid_queue_lock);
- if (midQ->mid_state == MID_REQUEST_SUBMITTED ||
- midQ->mid_state == MID_RESPONSE_RECEIVED) {
+ spin_lock(&midQ->mid_lock);
+ if (midQ->callback) {
/* no longer considered to be "in-flight" */
midQ->callback = release_mid;
- spin_unlock(&server->mid_queue_lock);
+ spin_unlock(&midQ->mid_lock);
add_credits(server, &credits, 0);
return rc;
}
- spin_unlock(&server->mid_queue_lock);
+ spin_unlock(&midQ->mid_lock);
}
rc = cifs_sync_mid_result(midQ, server);
rc = wait_for_response(server, midQ);
if (rc) {
send_cancel(server, &rqst, midQ);
- spin_lock(&server->mid_queue_lock);
- if (midQ->mid_state == MID_REQUEST_SUBMITTED ||
- midQ->mid_state == MID_RESPONSE_RECEIVED) {
+ spin_lock(&midQ->mid_lock);
+ if (midQ->callback) {
/* no longer considered to be "in-flight" */
midQ->callback = release_mid;
- spin_unlock(&server->mid_queue_lock);
+ spin_unlock(&midQ->mid_lock);
return rc;
}
- spin_unlock(&server->mid_queue_lock);
+ spin_unlock(&midQ->mid_lock);
}
/* We got the response - restart system call. */
cifs_dbg(FYI, "%s: issuing mid callbacks\n", __func__);
list_for_each_entry_safe(mid, nmid, &retry_list, qhead) {
list_del_init(&mid->qhead);
- mid->callback(mid);
+ mid_execute_callback(mid);
release_mid(mid);
}
list_del_init(&mid->qhead);
mid->mid_rc = mid_rc;
mid->mid_state = MID_RC;
- mid->callback(mid);
+ mid_execute_callback(mid);
release_mid(mid);
}
mid_entry = list_entry(tmp, struct mid_q_entry, qhead);
cifs_dbg(FYI, "Callback mid %llu\n", mid_entry->mid);
list_del_init(&mid_entry->qhead);
- mid_entry->callback(mid_entry);
+ mid_execute_callback(mid_entry);
release_mid(mid_entry);
}
/* 1/8th of sec is more than enough time for them to exit */
}
if (!mids[i]->multiRsp || mids[i]->multiEnd)
- mids[i]->callback(mids[i]);
+ mid_execute_callback(mids[i]);
release_mid(mids[i]);
} else if (server->ops->is_oplock_break &&
dw->server->ops->is_network_name_deleted(dw->buf,
dw->server);
- mid->callback(mid);
+ mid_execute_callback(mid);
} else {
spin_lock(&dw->server->srv_lock);
if (dw->server->tcpStatus == CifsNeedReconnect) {
mid->mid_state = MID_RETRY_NEEDED;
spin_unlock(&dw->server->mid_queue_lock);
spin_unlock(&dw->server->srv_lock);
- mid->callback(mid);
+ mid_execute_callback(mid);
} else {
spin_lock(&dw->server->mid_queue_lock);
mid->mid_state = MID_REQUEST_SUBMITTED;
temp = mempool_alloc(cifs_mid_poolp, GFP_NOFS);
memset(temp, 0, sizeof(struct mid_q_entry));
kref_init(&temp->refcount);
+ spin_lock_init(&temp->mid_lock);
temp->mid = le64_to_cpu(shdr->MessageId);
temp->credits = credits > 0 ? credits : 1;
temp->pid = current->pid;
cifs_server_dbg(FYI, "Cancelling wait for mid %llu cmd: %d\n",
midQ[i]->mid, le16_to_cpu(midQ[i]->command));
send_cancel(server, &rqst[i], midQ[i]);
- spin_lock(&server->mid_queue_lock);
+ spin_lock(&midQ[i]->mid_lock);
midQ[i]->wait_cancelled = true;
- if (midQ[i]->mid_state == MID_REQUEST_SUBMITTED ||
- midQ[i]->mid_state == MID_RESPONSE_RECEIVED) {
+ if (midQ[i]->callback) {
midQ[i]->callback = cifs_cancelled_callback;
cancelled_mid[i] = true;
credits[i].value = 0;
}
- spin_unlock(&server->mid_queue_lock);
+ spin_unlock(&midQ[i]->mid_lock);
}
}