mk-rootfs.sh

hc

2023-12-11 072de836f53be56a70cecf70b43ae43b7ce17376

 kernel/net/core/sysctl_net_core.c
..
..
@@ -22,9 +22,8 @@
22
22
 #include <net/busy_poll.h>
23
23
 #include <net/pkt_sched.h>
24
24
 
25
-static int zero = 0;
26
-static int one = 1;
27
-static int two __maybe_unused = 2;
25
+static int two = 2;
26
+static int three = 3;
28
27
 static int min_sndbuf = SOCK_MIN_SNDBUF;
29
28
 static int min_rcvbuf = SOCK_MIN_RCVBUF;
30
29
 static int max_skb_frags = MAX_SKB_FRAGS;
..
..
@@ -36,9 +35,19 @@
36
35
 int sysctl_fb_tunnels_only_for_init_net __read_mostly = 0;
37
36
 EXPORT_SYMBOL(sysctl_fb_tunnels_only_for_init_net);
38
37
 
38
+/* 0 - Keep current behavior:
39
+ *     IPv4: inherit all current settings from init_net
40
+ *     IPv6: reset all settings to default
41
+ * 1 - Both inherit all current settings from init_net
42
+ * 2 - Both reset all settings to default
43
+ * 3 - Both inherit all settings from current netns
44
+ */
45
+int sysctl_devconf_inherit_init_net __read_mostly;
46
+EXPORT_SYMBOL(sysctl_devconf_inherit_init_net);
47
+
39
48
 #ifdef CONFIG_RPS
40
49
 static int rps_sock_flow_sysctl(struct ctl_table *table, int write,
41
-				void __user *buffer, size_t *lenp, loff_t *ppos)
50
+				void *buffer, size_t *lenp, loff_t *ppos)
42
51
 {
43
52
 	unsigned int orig_size, size;
44
53
 	int ret, i;
..
..
@@ -86,12 +95,12 @@
86
95
 		if (sock_table != orig_sock_table) {
87
96
 			rcu_assign_pointer(rps_sock_flow_table, sock_table);
88
97
 			if (sock_table) {
89
-				static_key_slow_inc(&rps_needed);
90
-				static_key_slow_inc(&rfs_needed);
98
+				static_branch_inc(&rps_needed);
99
+				static_branch_inc(&rfs_needed);
91
100
 			}
92
101
 			if (orig_sock_table) {
93
-				static_key_slow_dec(&rps_needed);
94
-				static_key_slow_dec(&rfs_needed);
102
+				static_branch_dec(&rps_needed);
103
+				static_branch_dec(&rfs_needed);
95
104
 				synchronize_rcu();
96
105
 				vfree(orig_sock_table);
97
106
 			}
..
..
@@ -108,8 +117,7 @@
108
117
 static DEFINE_MUTEX(flow_limit_update_mutex);
109
118
 
110
119
 static int flow_limit_cpu_sysctl(struct ctl_table *table, int write,
111
-				 void __user *buffer, size_t *lenp,
112
-				 loff_t *ppos)
120
+				 void *buffer, size_t *lenp, loff_t *ppos)
113
121
 {
114
122
 	struct sd_flow_limit *cur;
115
123
 	struct softnet_data *sd;
..
..
@@ -120,7 +128,7 @@
120
128
 		return -ENOMEM;
121
129
 
122
130
 	if (write) {
123
-		ret = cpumask_parse_user(buffer, *lenp, mask);
131
+		ret = cpumask_parse(buffer, mask);
124
132
 		if (ret)
125
133
 			goto done;
126
134
 
..
..
@@ -173,10 +181,7 @@
173
181
 		}
174
182
 		if (len < *lenp)
175
183
 			kbuf[len++] = '\n';
176
-		if (copy_to_user(buffer, kbuf, len)) {
177
-			ret = -EFAULT;
178
-			goto done;
179
-		}
184
+		memcpy(buffer, kbuf, len);
180
185
 		*lenp = len;
181
186
 		*ppos += len;
182
187
 	}
..
..
@@ -187,8 +192,7 @@
187
192
 }
188
193
 
189
194
 static int flow_limit_table_len_sysctl(struct ctl_table *table, int write,
190
-				       void __user *buffer, size_t *lenp,
191
-				       loff_t *ppos)
195
+				       void *buffer, size_t *lenp, loff_t *ppos)
192
196
 {
193
197
 	unsigned int old, *ptr;
194
198
 	int ret;
..
..
@@ -210,7 +214,7 @@
210
214
 
211
215
 #ifdef CONFIG_NET_SCHED
212
216
 static int set_default_qdisc(struct ctl_table *table, int write,
213
-			     void __user *buffer, size_t *lenp, loff_t *ppos)
217
+			     void *buffer, size_t *lenp, loff_t *ppos)
214
218
 {
215
219
 	char id[IFNAMSIZ];
216
220
 	struct ctl_table tbl = {
..
..
@@ -229,23 +233,25 @@
229
233
 #endif
230
234
 
231
235
 static int proc_do_dev_weight(struct ctl_table *table, int write,
232
-			   void __user *buffer, size_t *lenp, loff_t *ppos)
236
+			   void *buffer, size_t *lenp, loff_t *ppos)
233
237
 {
234
-	int ret;
238
+	static DEFINE_MUTEX(dev_weight_mutex);
239
+	int ret, weight;
235
240
 
241
+	mutex_lock(&dev_weight_mutex);
236
242
 	ret = proc_dointvec(table, write, buffer, lenp, ppos);
237
-	if (ret != 0)
238
-		return ret;
239
-
240
-	dev_rx_weight = weight_p * dev_weight_rx_bias;
241
-	dev_tx_weight = weight_p * dev_weight_tx_bias;
243
+	if (!ret && write) {
244
+		weight = READ_ONCE(weight_p);
245
+		WRITE_ONCE(dev_rx_weight, weight * dev_weight_rx_bias);
246
+		WRITE_ONCE(dev_tx_weight, weight * dev_weight_tx_bias);
247
+	}
248
+	mutex_unlock(&dev_weight_mutex);
242
249
 
243
250
 	return ret;
244
251
 }
245
252
 
246
-#ifdef CONFIG_ETHTOOL
247
253
 static int proc_do_rss_key(struct ctl_table *table, int write,
248
-			   void __user *buffer, size_t *lenp, loff_t *ppos)
254
+			   void *buffer, size_t *lenp, loff_t *ppos)
249
255
 {
250
256
 	struct ctl_table fake_table;
251
257
 	char buf[NETDEV_RSS_KEY_LEN * 3];
..
..
@@ -255,11 +261,10 @@
255
261
 	fake_table.maxlen = sizeof(buf);
256
262
 	return proc_dostring(&fake_table, write, buffer, lenp, ppos);
257
263
 }
258
-#endif
259
264
 
260
265
 #ifdef CONFIG_BPF_JIT
261
266
 static int proc_dointvec_minmax_bpf_enable(struct ctl_table *table, int write,
262
-					   void __user *buffer, size_t *lenp,
267
+					   void *buffer, size_t *lenp,
263
268
 					   loff_t *ppos)
264
269
 {
265
270
 	int ret, jit_enable = *(int *)table->data;
..
..
@@ -286,8 +291,7 @@
286
291
 # ifdef CONFIG_HAVE_EBPF_JIT
287
292
 static int
288
293
 proc_dointvec_minmax_bpf_restricted(struct ctl_table *table, int write,
289
-				    void __user *buffer, size_t *lenp,
290
-				    loff_t *ppos)
294
+				    void *buffer, size_t *lenp, loff_t *ppos)
291
295
 {
292
296
 	if (!capable(CAP_SYS_ADMIN))
293
297
 		return -EPERM;
..
..
@@ -298,8 +302,7 @@
298
302
 
299
303
 static int
300
304
 proc_dolongvec_minmax_bpf_restricted(struct ctl_table *table, int write,
301
-				     void __user *buffer, size_t *lenp,
302
-				     loff_t *ppos)
305
+				     void *buffer, size_t *lenp, loff_t *ppos)
303
306
 {
304
307
 	if (!capable(CAP_SYS_ADMIN))
305
308
 		return -EPERM;
..
..
@@ -370,7 +373,6 @@
370
373
 		.mode		= 0644,
371
374
 		.proc_handler	= proc_dointvec
372
375
 	},
373
-#ifdef CONFIG_ETHTOOL
374
376
 	{
375
377
 		.procname	= "netdev_rss_key",
376
378
 		.data		= &netdev_rss_key,
..
..
@@ -378,7 +380,6 @@
378
380
 		.mode		= 0444,
379
381
 		.proc_handler	= proc_do_rss_key,
380
382
 	},
381
-#endif
382
383
 #ifdef CONFIG_BPF_JIT
383
384
 	{
384
385
 		.procname	= "bpf_jit_enable",
..
..
@@ -387,10 +388,10 @@
387
388
 		.mode		= 0644,
388
389
 		.proc_handler	= proc_dointvec_minmax_bpf_enable,
389
390
 # ifdef CONFIG_BPF_JIT_ALWAYS_ON
390
-		.extra1		= &one,
391
-		.extra2		= &one,
391
+		.extra1		= SYSCTL_ONE,
392
+		.extra2		= SYSCTL_ONE,
392
393
 # else
393
-		.extra1		= &zero,
394
+		.extra1		= SYSCTL_ZERO,
394
395
 		.extra2		= &two,
395
396
 # endif
396
397
 	},
..
..
@@ -401,7 +402,7 @@
401
402
 		.maxlen		= sizeof(int),
402
403
 		.mode		= 0600,
403
404
 		.proc_handler	= proc_dointvec_minmax_bpf_restricted,
404
-		.extra1		= &zero,
405
+		.extra1		= SYSCTL_ZERO,
405
406
 		.extra2		= &two,
406
407
 	},
407
408
 	{
..
..
@@ -410,8 +411,8 @@
410
411
 		.maxlen		= sizeof(int),
411
412
 		.mode		= 0600,
412
413
 		.proc_handler	= proc_dointvec_minmax_bpf_restricted,
413
-		.extra1		= &zero,
414
-		.extra2		= &one,
414
+		.extra1		= SYSCTL_ZERO,
415
+		.extra2		= SYSCTL_ONE,
415
416
 	},
416
417
 # endif
417
418
 	{
..
..
@@ -458,8 +459,8 @@
458
459
 		.maxlen		= sizeof(int),
459
460
 		.mode		= 0644,
460
461
 		.proc_handler	= proc_dointvec_minmax,
461
-		.extra1		= &zero,
462
-		.extra2		= &one
462
+		.extra1		= SYSCTL_ZERO,
463
+		.extra2		= SYSCTL_ONE
463
464
 	},
464
465
 #ifdef CONFIG_RPS
465
466
 	{
..
..
@@ -490,7 +491,7 @@
490
491
 		.maxlen		= sizeof(unsigned int),
491
492
 		.mode		= 0644,
492
493
 		.proc_handler	= proc_dointvec_minmax,
493
-		.extra1		= &zero,
494
+		.extra1		= SYSCTL_ZERO,
494
495
 	},
495
496
 	{
496
497
 		.procname	= "busy_read",
..
..
@@ -498,7 +499,7 @@
498
499
 		.maxlen		= sizeof(unsigned int),
499
500
 		.mode		= 0644,
500
501
 		.proc_handler	= proc_dointvec_minmax,
501
-		.extra1		= &zero,
502
+		.extra1		= SYSCTL_ZERO,
502
503
 	},
503
504
 #endif
504
505
 #ifdef CONFIG_NET_SCHED
..
..
@@ -530,7 +531,7 @@
530
531
 		.maxlen		= sizeof(int),
531
532
 		.mode		= 0644,
532
533
 		.proc_handler	= proc_dointvec_minmax,
533
-		.extra1		= &one,
534
+		.extra1		= SYSCTL_ONE,
534
535
 		.extra2		= &max_skb_frags,
535
536
 	},
536
537
 	{
..
..
@@ -539,7 +540,7 @@
539
540
 		.maxlen		= sizeof(unsigned int),
540
541
 		.mode		= 0644,
541
542
 		.proc_handler	= proc_dointvec_minmax,
542
-		.extra1		= &zero,
543
+		.extra1		= SYSCTL_ZERO,
543
544
 	},
544
545
 	{
545
546
 		.procname	= "fb_tunnels_only_for_init_net",
..
..
@@ -547,8 +548,32 @@
547
548
 		.maxlen		= sizeof(int),
548
549
 		.mode		= 0644,
549
550
 		.proc_handler	= proc_dointvec_minmax,
550
-		.extra1		= &zero,
551
-		.extra2		= &one,
551
+		.extra1		= SYSCTL_ZERO,
552
+		.extra2		= &two,
553
+	},
554
+	{
555
+		.procname	= "devconf_inherit_init_net",
556
+		.data		= &sysctl_devconf_inherit_init_net,
557
+		.maxlen		= sizeof(int),
558
+		.mode		= 0644,
559
+		.proc_handler	= proc_dointvec_minmax,
560
+		.extra1		= SYSCTL_ZERO,
561
+		.extra2		= &three,
562
+	},
563
+	{
564
+		.procname	= "high_order_alloc_disable",
565
+		.data		= &net_high_order_alloc_disable_key.key,
566
+		.maxlen         = sizeof(net_high_order_alloc_disable_key),
567
+		.mode		= 0644,
568
+		.proc_handler	= proc_do_static_key,
569
+	},
570
+	{
571
+		.procname	= "gro_normal_batch",
572
+		.data		= &gro_normal_batch,
573
+		.maxlen		= sizeof(unsigned int),
574
+		.mode		= 0644,
575
+		.proc_handler	= proc_dointvec_minmax,
576
+		.extra1		= SYSCTL_ONE,
552
577
 	},
553
578
 	{ }
554
579
 };
..
..
@@ -559,12 +584,25 @@
559
584
 		.data		= &init_net.core.sysctl_somaxconn,
560
585
 		.maxlen		= sizeof(int),
561
586
 		.mode		= 0644,
562
-		.extra1		= &zero,
587
+		.extra1		= SYSCTL_ZERO,
563
588
 		.proc_handler	= proc_dointvec_minmax
564
589
 	},
565
590
 	{ }
566
591
 };
567
592
 
593
+static int __init fb_tunnels_only_for_init_net_sysctl_setup(char *str)
594
+{
595
+	/* fallback tunnels for initns only */
596
+	if (!strncmp(str, "initns", 6))
597
+		sysctl_fb_tunnels_only_for_init_net = 1;
598
+	/* no fallback tunnels anywhere */
599
+	else if (!strncmp(str, "none", 4))
600
+		sysctl_fb_tunnels_only_for_init_net = 2;
601
+
602
+	return 1;
603
+}
604
+__setup("fb_tunnels=", fb_tunnels_only_for_init_net_sysctl_setup);
605
+
568
606
 static __net_init int sysctl_core_net_init(struct net *net)
569
607
 {
570
608
 	struct ctl_table *tbl;