]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* File veth.c created by Kyle A. Lucke on Mon Aug 7 2000. */ |
2 | /* | |
3 | * IBM eServer iSeries Virtual Ethernet Device Driver | |
4 | * Copyright (C) 2001 Kyle A. Lucke (klucke@us.ibm.com), IBM Corp. | |
5 | * Substantially cleaned up by: | |
6 | * Copyright (C) 2003 David Gibson <dwg@au1.ibm.com>, IBM Corporation. | |
269cd381 | 7 | * Copyright (C) 2004-2005 Michael Ellerman, IBM Corporation. |
1da177e4 LT |
8 | * |
9 | * This program is free software; you can redistribute it and/or | |
10 | * modify it under the terms of the GNU General Public License as | |
11 | * published by the Free Software Foundation; either version 2 of the | |
12 | * License, or (at your option) any later version. | |
13 | * | |
14 | * This program is distributed in the hope that it will be useful, but | |
15 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
16 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
17 | * General Public License for more details. | |
18 | * | |
19 | * You should have received a copy of the GNU General Public License | |
20 | * along with this program; if not, write to the Free Software | |
21 | * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 | |
22 | * USA | |
23 | * | |
24 | * | |
25 | * This module implements the virtual ethernet device for iSeries LPAR | |
26 | * Linux. It uses hypervisor message passing to implement an | |
27 | * ethernet-like network device communicating between partitions on | |
28 | * the iSeries. | |
29 | * | |
30 | * The iSeries LPAR hypervisor currently allows for up to 16 different | |
31 | * virtual ethernets. These are all dynamically configurable on | |
32 | * OS/400 partitions, but dynamic configuration is not supported under | |
33 | * Linux yet. An ethXX network device will be created for each | |
34 | * virtual ethernet this partition is connected to. | |
35 | * | |
36 | * - This driver is responsible for routing packets to and from other | |
37 | * partitions. The MAC addresses used by the virtual ethernets | |
38 | * contains meaning and must not be modified. | |
39 | * | |
40 | * - Having 2 virtual ethernets to the same remote partition DOES NOT | |
41 | * double the available bandwidth. The 2 devices will share the | |
42 | * available hypervisor bandwidth. | |
43 | * | |
44 | * - If you send a packet to your own mac address, it will just be | |
45 | * dropped, you won't get it on the receive side. | |
46 | * | |
47 | * - Multicast is implemented by sending the frame frame to every | |
48 | * other partition. It is the responsibility of the receiving | |
49 | * partition to filter the addresses desired. | |
50 | * | |
51 | * Tunable parameters: | |
52 | * | |
53 | * VETH_NUMBUFFERS: This compile time option defaults to 120. It | |
54 | * controls how much memory Linux will allocate per remote partition | |
55 | * it is communicating with. It can be thought of as the maximum | |
56 | * number of packets outstanding to a remote partition at a time. | |
57 | */ | |
58 | ||
1da177e4 | 59 | #include <linux/module.h> |
1da177e4 LT |
60 | #include <linux/types.h> |
61 | #include <linux/errno.h> | |
62 | #include <linux/ioport.h> | |
63 | #include <linux/kernel.h> | |
64 | #include <linux/netdevice.h> | |
65 | #include <linux/etherdevice.h> | |
66 | #include <linux/skbuff.h> | |
67 | #include <linux/init.h> | |
68 | #include <linux/delay.h> | |
69 | #include <linux/mm.h> | |
70 | #include <linux/ethtool.h> | |
07fb3f45 | 71 | #include <linux/if_ether.h> |
5a0e3ad6 | 72 | #include <linux/slab.h> |
426c1a11 SR |
73 | |
74 | #include <asm/abs_addr.h> | |
bbc8b628 | 75 | #include <asm/iseries/mf.h> |
1da177e4 | 76 | #include <asm/uaccess.h> |
687d18ab | 77 | #include <asm/firmware.h> |
15b17189 | 78 | #include <asm/iseries/hv_lp_config.h> |
1ec65d76 | 79 | #include <asm/iseries/hv_types.h> |
e45423ea | 80 | #include <asm/iseries/hv_lp_event.h> |
1da177e4 LT |
81 | #include <asm/iommu.h> |
82 | #include <asm/vio.h> | |
83 | ||
61a3c696 ME |
84 | #undef DEBUG |
85 | ||
1da177e4 LT |
86 | MODULE_AUTHOR("Kyle Lucke <klucke@us.ibm.com>"); |
87 | MODULE_DESCRIPTION("iSeries Virtual ethernet driver"); | |
88 | MODULE_LICENSE("GPL"); | |
89 | ||
59f17aeb ME |
90 | #define VETH_EVENT_CAP (0) |
91 | #define VETH_EVENT_FRAMES (1) | |
92 | #define VETH_EVENT_MONITOR (2) | |
93 | #define VETH_EVENT_FRAMES_ACK (3) | |
642d1a4c ME |
94 | |
95 | #define VETH_MAX_ACKS_PER_MSG (20) | |
96 | #define VETH_MAX_FRAMES_PER_MSG (6) | |
97 | ||
59f17aeb | 98 | struct veth_frames_data { |
642d1a4c ME |
99 | u32 addr[VETH_MAX_FRAMES_PER_MSG]; |
100 | u16 len[VETH_MAX_FRAMES_PER_MSG]; | |
101 | u32 eofmask; | |
102 | }; | |
103 | #define VETH_EOF_SHIFT (32-VETH_MAX_FRAMES_PER_MSG) | |
104 | ||
59f17aeb | 105 | struct veth_frames_ack_data { |
642d1a4c ME |
106 | u16 token[VETH_MAX_ACKS_PER_MSG]; |
107 | }; | |
108 | ||
59f17aeb | 109 | struct veth_cap_data { |
642d1a4c ME |
110 | u8 caps_version; |
111 | u8 rsvd1; | |
112 | u16 num_buffers; | |
113 | u16 ack_threshold; | |
114 | u16 rsvd2; | |
115 | u32 ack_timeout; | |
116 | u32 rsvd3; | |
117 | u64 rsvd4[3]; | |
118 | }; | |
119 | ||
59f17aeb | 120 | struct veth_lpevent { |
642d1a4c ME |
121 | struct HvLpEvent base_event; |
122 | union { | |
59f17aeb ME |
123 | struct veth_cap_data caps_data; |
124 | struct veth_frames_data frames_data; | |
125 | struct veth_frames_ack_data frames_ack_data; | |
642d1a4c ME |
126 | } u; |
127 | ||
128 | }; | |
129 | ||
ee05f031 ME |
130 | #define DRV_NAME "iseries_veth" |
131 | #define DRV_VERSION "2.0" | |
132 | ||
1da177e4 LT |
133 | #define VETH_NUMBUFFERS (120) |
134 | #define VETH_ACKTIMEOUT (1000000) /* microseconds */ | |
135 | #define VETH_MAX_MCAST (12) | |
136 | ||
137 | #define VETH_MAX_MTU (9000) | |
138 | ||
139 | #if VETH_NUMBUFFERS < 10 | |
140 | #define ACK_THRESHOLD (1) | |
141 | #elif VETH_NUMBUFFERS < 20 | |
142 | #define ACK_THRESHOLD (4) | |
143 | #elif VETH_NUMBUFFERS < 40 | |
144 | #define ACK_THRESHOLD (10) | |
145 | #else | |
146 | #define ACK_THRESHOLD (20) | |
147 | #endif | |
148 | ||
149 | #define VETH_STATE_SHUTDOWN (0x0001) | |
150 | #define VETH_STATE_OPEN (0x0002) | |
151 | #define VETH_STATE_RESET (0x0004) | |
152 | #define VETH_STATE_SENTMON (0x0008) | |
153 | #define VETH_STATE_SENTCAPS (0x0010) | |
154 | #define VETH_STATE_GOTCAPACK (0x0020) | |
155 | #define VETH_STATE_GOTCAPS (0x0040) | |
156 | #define VETH_STATE_SENTCAPACK (0x0080) | |
157 | #define VETH_STATE_READY (0x0100) | |
158 | ||
159 | struct veth_msg { | |
160 | struct veth_msg *next; | |
59f17aeb | 161 | struct veth_frames_data data; |
1da177e4 | 162 | int token; |
b08bd5c0 | 163 | int in_use; |
1da177e4 LT |
164 | struct sk_buff *skb; |
165 | struct device *dev; | |
166 | }; | |
167 | ||
168 | struct veth_lpar_connection { | |
169 | HvLpIndex remote_lp; | |
c4028958 | 170 | struct delayed_work statemachine_wq; |
1da177e4 LT |
171 | struct veth_msg *msgs; |
172 | int num_events; | |
59f17aeb | 173 | struct veth_cap_data local_caps; |
1da177e4 | 174 | |
f0c129ca | 175 | struct kobject kobject; |
1da177e4 LT |
176 | struct timer_list ack_timer; |
177 | ||
24562ffa ME |
178 | struct timer_list reset_timer; |
179 | unsigned int reset_timeout; | |
180 | unsigned long last_contact; | |
181 | int outstanding_tx; | |
182 | ||
1da177e4 LT |
183 | spinlock_t lock; |
184 | unsigned long state; | |
185 | HvLpInstanceId src_inst; | |
186 | HvLpInstanceId dst_inst; | |
59f17aeb | 187 | struct veth_lpevent cap_event, cap_ack_event; |
1da177e4 LT |
188 | u16 pending_acks[VETH_MAX_ACKS_PER_MSG]; |
189 | u32 num_pending_acks; | |
190 | ||
191 | int num_ack_events; | |
59f17aeb | 192 | struct veth_cap_data remote_caps; |
1da177e4 LT |
193 | u32 ack_timeout; |
194 | ||
1da177e4 LT |
195 | struct veth_msg *msg_stack_head; |
196 | }; | |
197 | ||
198 | struct veth_port { | |
199 | struct device *dev; | |
1da177e4 LT |
200 | u64 mac_addr; |
201 | HvLpIndexMap lpar_map; | |
202 | ||
e0808494 ME |
203 | /* queue_lock protects the stopped_map and dev's queue. */ |
204 | spinlock_t queue_lock; | |
205 | HvLpIndexMap stopped_map; | |
1da177e4 | 206 | |
e0808494 | 207 | /* mcast_gate protects promiscuous, num_mcast & mcast_addr. */ |
1da177e4 LT |
208 | rwlock_t mcast_gate; |
209 | int promiscuous; | |
1da177e4 LT |
210 | int num_mcast; |
211 | u64 mcast_addr[VETH_MAX_MCAST]; | |
07a5c172 ME |
212 | |
213 | struct kobject kobject; | |
1da177e4 LT |
214 | }; |
215 | ||
216 | static HvLpIndex this_lp; | |
217 | static struct veth_lpar_connection *veth_cnx[HVMAXARCHITECTEDLPS]; /* = 0 */ | |
218 | static struct net_device *veth_dev[HVMAXARCHITECTEDVIRTUALLANS]; /* = 0 */ | |
219 | ||
220 | static int veth_start_xmit(struct sk_buff *skb, struct net_device *dev); | |
221 | static void veth_recycle_msg(struct veth_lpar_connection *, struct veth_msg *); | |
e0808494 ME |
222 | static void veth_wake_queues(struct veth_lpar_connection *cnx); |
223 | static void veth_stop_queues(struct veth_lpar_connection *cnx); | |
59f17aeb | 224 | static void veth_receive(struct veth_lpar_connection *, struct veth_lpevent *); |
f0c129ca | 225 | static void veth_release_connection(struct kobject *kobject); |
24562ffa ME |
226 | static void veth_timed_ack(unsigned long ptr); |
227 | static void veth_timed_reset(unsigned long ptr); | |
f0c129ca | 228 | |
1da177e4 LT |
229 | /* |
230 | * Utility functions | |
231 | */ | |
232 | ||
61a3c696 | 233 | #define veth_info(fmt, args...) \ |
ee05f031 | 234 | printk(KERN_INFO DRV_NAME ": " fmt, ## args) |
1da177e4 LT |
235 | |
236 | #define veth_error(fmt, args...) \ | |
ee05f031 | 237 | printk(KERN_ERR DRV_NAME ": Error: " fmt, ## args) |
61a3c696 ME |
238 | |
239 | #ifdef DEBUG | |
240 | #define veth_debug(fmt, args...) \ | |
ee05f031 | 241 | printk(KERN_DEBUG DRV_NAME ": " fmt, ## args) |
61a3c696 ME |
242 | #else |
243 | #define veth_debug(fmt, args...) do {} while (0) | |
244 | #endif | |
1da177e4 | 245 | |
d7893ddd | 246 | /* You must hold the connection's lock when you call this function. */ |
1da177e4 LT |
247 | static inline void veth_stack_push(struct veth_lpar_connection *cnx, |
248 | struct veth_msg *msg) | |
249 | { | |
1da177e4 LT |
250 | msg->next = cnx->msg_stack_head; |
251 | cnx->msg_stack_head = msg; | |
1da177e4 LT |
252 | } |
253 | ||
d7893ddd | 254 | /* You must hold the connection's lock when you call this function. */ |
1da177e4 LT |
255 | static inline struct veth_msg *veth_stack_pop(struct veth_lpar_connection *cnx) |
256 | { | |
1da177e4 LT |
257 | struct veth_msg *msg; |
258 | ||
1da177e4 LT |
259 | msg = cnx->msg_stack_head; |
260 | if (msg) | |
261 | cnx->msg_stack_head = cnx->msg_stack_head->next; | |
d7893ddd | 262 | |
1da177e4 LT |
263 | return msg; |
264 | } | |
265 | ||
e0808494 ME |
266 | /* You must hold the connection's lock when you call this function. */ |
267 | static inline int veth_stack_is_empty(struct veth_lpar_connection *cnx) | |
268 | { | |
269 | return cnx->msg_stack_head == NULL; | |
270 | } | |
271 | ||
1da177e4 LT |
272 | static inline HvLpEvent_Rc |
273 | veth_signalevent(struct veth_lpar_connection *cnx, u16 subtype, | |
274 | HvLpEvent_AckInd ackind, HvLpEvent_AckType acktype, | |
275 | u64 token, | |
276 | u64 data1, u64 data2, u64 data3, u64 data4, u64 data5) | |
277 | { | |
278 | return HvCallEvent_signalLpEventFast(cnx->remote_lp, | |
279 | HvLpEvent_Type_VirtualLan, | |
280 | subtype, ackind, acktype, | |
281 | cnx->src_inst, | |
282 | cnx->dst_inst, | |
283 | token, data1, data2, data3, | |
284 | data4, data5); | |
285 | } | |
286 | ||
287 | static inline HvLpEvent_Rc veth_signaldata(struct veth_lpar_connection *cnx, | |
288 | u16 subtype, u64 token, void *data) | |
289 | { | |
290 | u64 *p = (u64 *) data; | |
291 | ||
292 | return veth_signalevent(cnx, subtype, HvLpEvent_AckInd_NoAck, | |
293 | HvLpEvent_AckType_ImmediateAck, | |
294 | token, p[0], p[1], p[2], p[3], p[4]); | |
295 | } | |
296 | ||
297 | struct veth_allocation { | |
298 | struct completion c; | |
299 | int num; | |
300 | }; | |
301 | ||
302 | static void veth_complete_allocation(void *parm, int number) | |
303 | { | |
304 | struct veth_allocation *vc = (struct veth_allocation *)parm; | |
305 | ||
306 | vc->num = number; | |
307 | complete(&vc->c); | |
308 | } | |
309 | ||
310 | static int veth_allocate_events(HvLpIndex rlp, int number) | |
311 | { | |
a2879fef BH |
312 | struct veth_allocation vc = |
313 | { COMPLETION_INITIALIZER_ONSTACK(vc.c), 0 }; | |
1da177e4 LT |
314 | |
315 | mf_allocate_lp_events(rlp, HvLpEvent_Type_VirtualLan, | |
59f17aeb | 316 | sizeof(struct veth_lpevent), number, |
1da177e4 LT |
317 | &veth_complete_allocation, &vc); |
318 | wait_for_completion(&vc.c); | |
319 | ||
320 | return vc.num; | |
321 | } | |
322 | ||
76812d81 ME |
323 | /* |
324 | * sysfs support | |
325 | */ | |
326 | ||
327 | struct veth_cnx_attribute { | |
328 | struct attribute attr; | |
329 | ssize_t (*show)(struct veth_lpar_connection *, char *buf); | |
330 | ssize_t (*store)(struct veth_lpar_connection *, const char *buf); | |
331 | }; | |
332 | ||
333 | static ssize_t veth_cnx_attribute_show(struct kobject *kobj, | |
334 | struct attribute *attr, char *buf) | |
335 | { | |
336 | struct veth_cnx_attribute *cnx_attr; | |
337 | struct veth_lpar_connection *cnx; | |
338 | ||
339 | cnx_attr = container_of(attr, struct veth_cnx_attribute, attr); | |
340 | cnx = container_of(kobj, struct veth_lpar_connection, kobject); | |
341 | ||
342 | if (!cnx_attr->show) | |
343 | return -EIO; | |
344 | ||
345 | return cnx_attr->show(cnx, buf); | |
346 | } | |
347 | ||
348 | #define CUSTOM_CNX_ATTR(_name, _format, _expression) \ | |
349 | static ssize_t _name##_show(struct veth_lpar_connection *cnx, char *buf)\ | |
350 | { \ | |
351 | return sprintf(buf, _format, _expression); \ | |
352 | } \ | |
353 | struct veth_cnx_attribute veth_cnx_attr_##_name = __ATTR_RO(_name) | |
354 | ||
355 | #define SIMPLE_CNX_ATTR(_name) \ | |
356 | CUSTOM_CNX_ATTR(_name, "%lu\n", (unsigned long)cnx->_name) | |
357 | ||
358 | SIMPLE_CNX_ATTR(outstanding_tx); | |
359 | SIMPLE_CNX_ATTR(remote_lp); | |
360 | SIMPLE_CNX_ATTR(num_events); | |
361 | SIMPLE_CNX_ATTR(src_inst); | |
362 | SIMPLE_CNX_ATTR(dst_inst); | |
363 | SIMPLE_CNX_ATTR(num_pending_acks); | |
364 | SIMPLE_CNX_ATTR(num_ack_events); | |
365 | CUSTOM_CNX_ATTR(ack_timeout, "%d\n", jiffies_to_msecs(cnx->ack_timeout)); | |
366 | CUSTOM_CNX_ATTR(reset_timeout, "%d\n", jiffies_to_msecs(cnx->reset_timeout)); | |
367 | CUSTOM_CNX_ATTR(state, "0x%.4lX\n", cnx->state); | |
368 | CUSTOM_CNX_ATTR(last_contact, "%d\n", cnx->last_contact ? | |
369 | jiffies_to_msecs(jiffies - cnx->last_contact) : 0); | |
370 | ||
371 | #define GET_CNX_ATTR(_name) (&veth_cnx_attr_##_name.attr) | |
372 | ||
373 | static struct attribute *veth_cnx_default_attrs[] = { | |
374 | GET_CNX_ATTR(outstanding_tx), | |
375 | GET_CNX_ATTR(remote_lp), | |
376 | GET_CNX_ATTR(num_events), | |
377 | GET_CNX_ATTR(reset_timeout), | |
378 | GET_CNX_ATTR(last_contact), | |
379 | GET_CNX_ATTR(state), | |
380 | GET_CNX_ATTR(src_inst), | |
381 | GET_CNX_ATTR(dst_inst), | |
382 | GET_CNX_ATTR(num_pending_acks), | |
383 | GET_CNX_ATTR(num_ack_events), | |
384 | GET_CNX_ATTR(ack_timeout), | |
385 | NULL | |
386 | }; | |
387 | ||
52cf25d0 | 388 | static const struct sysfs_ops veth_cnx_sysfs_ops = { |
76812d81 ME |
389 | .show = veth_cnx_attribute_show |
390 | }; | |
391 | ||
392 | static struct kobj_type veth_lpar_connection_ktype = { | |
393 | .release = veth_release_connection, | |
394 | .sysfs_ops = &veth_cnx_sysfs_ops, | |
395 | .default_attrs = veth_cnx_default_attrs | |
396 | }; | |
397 | ||
07a5c172 ME |
398 | struct veth_port_attribute { |
399 | struct attribute attr; | |
400 | ssize_t (*show)(struct veth_port *, char *buf); | |
401 | ssize_t (*store)(struct veth_port *, const char *buf); | |
402 | }; | |
403 | ||
404 | static ssize_t veth_port_attribute_show(struct kobject *kobj, | |
405 | struct attribute *attr, char *buf) | |
406 | { | |
407 | struct veth_port_attribute *port_attr; | |
408 | struct veth_port *port; | |
409 | ||
410 | port_attr = container_of(attr, struct veth_port_attribute, attr); | |
411 | port = container_of(kobj, struct veth_port, kobject); | |
412 | ||
413 | if (!port_attr->show) | |
414 | return -EIO; | |
415 | ||
416 | return port_attr->show(port, buf); | |
417 | } | |
418 | ||
419 | #define CUSTOM_PORT_ATTR(_name, _format, _expression) \ | |
420 | static ssize_t _name##_show(struct veth_port *port, char *buf) \ | |
421 | { \ | |
422 | return sprintf(buf, _format, _expression); \ | |
423 | } \ | |
424 | struct veth_port_attribute veth_port_attr_##_name = __ATTR_RO(_name) | |
425 | ||
426 | #define SIMPLE_PORT_ATTR(_name) \ | |
427 | CUSTOM_PORT_ATTR(_name, "%lu\n", (unsigned long)port->_name) | |
428 | ||
429 | SIMPLE_PORT_ATTR(promiscuous); | |
430 | SIMPLE_PORT_ATTR(num_mcast); | |
431 | CUSTOM_PORT_ATTR(lpar_map, "0x%X\n", port->lpar_map); | |
432 | CUSTOM_PORT_ATTR(stopped_map, "0x%X\n", port->stopped_map); | |
a1c5a893 | 433 | CUSTOM_PORT_ATTR(mac_addr, "0x%llX\n", port->mac_addr); |
07a5c172 ME |
434 | |
435 | #define GET_PORT_ATTR(_name) (&veth_port_attr_##_name.attr) | |
436 | static struct attribute *veth_port_default_attrs[] = { | |
437 | GET_PORT_ATTR(mac_addr), | |
438 | GET_PORT_ATTR(lpar_map), | |
439 | GET_PORT_ATTR(stopped_map), | |
440 | GET_PORT_ATTR(promiscuous), | |
441 | GET_PORT_ATTR(num_mcast), | |
442 | NULL | |
443 | }; | |
444 | ||
52cf25d0 | 445 | static const struct sysfs_ops veth_port_sysfs_ops = { |
07a5c172 ME |
446 | .show = veth_port_attribute_show |
447 | }; | |
448 | ||
449 | static struct kobj_type veth_port_ktype = { | |
450 | .sysfs_ops = &veth_port_sysfs_ops, | |
451 | .default_attrs = veth_port_default_attrs | |
452 | }; | |
453 | ||
1da177e4 LT |
454 | /* |
455 | * LPAR connection code | |
456 | */ | |
457 | ||
458 | static inline void veth_kick_statemachine(struct veth_lpar_connection *cnx) | |
459 | { | |
c4028958 | 460 | schedule_delayed_work(&cnx->statemachine_wq, 0); |
1da177e4 LT |
461 | } |
462 | ||
463 | static void veth_take_cap(struct veth_lpar_connection *cnx, | |
59f17aeb | 464 | struct veth_lpevent *event) |
1da177e4 LT |
465 | { |
466 | unsigned long flags; | |
467 | ||
468 | spin_lock_irqsave(&cnx->lock, flags); | |
469 | /* Receiving caps may mean the other end has just come up, so | |
470 | * we need to reload the instance ID of the far end */ | |
471 | cnx->dst_inst = | |
472 | HvCallEvent_getTargetLpInstanceId(cnx->remote_lp, | |
473 | HvLpEvent_Type_VirtualLan); | |
474 | ||
475 | if (cnx->state & VETH_STATE_GOTCAPS) { | |
61a3c696 | 476 | veth_error("Received a second capabilities from LPAR %d.\n", |
1da177e4 LT |
477 | cnx->remote_lp); |
478 | event->base_event.xRc = HvLpEvent_Rc_BufferNotAvailable; | |
479 | HvCallEvent_ackLpEvent((struct HvLpEvent *) event); | |
480 | } else { | |
481 | memcpy(&cnx->cap_event, event, sizeof(cnx->cap_event)); | |
482 | cnx->state |= VETH_STATE_GOTCAPS; | |
483 | veth_kick_statemachine(cnx); | |
484 | } | |
485 | spin_unlock_irqrestore(&cnx->lock, flags); | |
486 | } | |
487 | ||
488 | static void veth_take_cap_ack(struct veth_lpar_connection *cnx, | |
59f17aeb | 489 | struct veth_lpevent *event) |
1da177e4 LT |
490 | { |
491 | unsigned long flags; | |
492 | ||
493 | spin_lock_irqsave(&cnx->lock, flags); | |
494 | if (cnx->state & VETH_STATE_GOTCAPACK) { | |
61a3c696 | 495 | veth_error("Received a second capabilities ack from LPAR %d.\n", |
1da177e4 LT |
496 | cnx->remote_lp); |
497 | } else { | |
498 | memcpy(&cnx->cap_ack_event, event, | |
b607bd90 | 499 | sizeof(cnx->cap_ack_event)); |
1da177e4 LT |
500 | cnx->state |= VETH_STATE_GOTCAPACK; |
501 | veth_kick_statemachine(cnx); | |
502 | } | |
503 | spin_unlock_irqrestore(&cnx->lock, flags); | |
504 | } | |
505 | ||
506 | static void veth_take_monitor_ack(struct veth_lpar_connection *cnx, | |
59f17aeb | 507 | struct veth_lpevent *event) |
1da177e4 LT |
508 | { |
509 | unsigned long flags; | |
510 | ||
511 | spin_lock_irqsave(&cnx->lock, flags); | |
61a3c696 | 512 | veth_debug("cnx %d: lost connection.\n", cnx->remote_lp); |
58c5900b ME |
513 | |
514 | /* Avoid kicking the statemachine once we're shutdown. | |
515 | * It's unnecessary and it could break veth_stop_connection(). */ | |
516 | ||
517 | if (! (cnx->state & VETH_STATE_SHUTDOWN)) { | |
518 | cnx->state |= VETH_STATE_RESET; | |
519 | veth_kick_statemachine(cnx); | |
520 | } | |
1da177e4 LT |
521 | spin_unlock_irqrestore(&cnx->lock, flags); |
522 | } | |
523 | ||
59f17aeb | 524 | static void veth_handle_ack(struct veth_lpevent *event) |
1da177e4 LT |
525 | { |
526 | HvLpIndex rlp = event->base_event.xTargetLp; | |
527 | struct veth_lpar_connection *cnx = veth_cnx[rlp]; | |
528 | ||
529 | BUG_ON(! cnx); | |
530 | ||
531 | switch (event->base_event.xSubtype) { | |
59f17aeb | 532 | case VETH_EVENT_CAP: |
1da177e4 LT |
533 | veth_take_cap_ack(cnx, event); |
534 | break; | |
59f17aeb | 535 | case VETH_EVENT_MONITOR: |
1da177e4 LT |
536 | veth_take_monitor_ack(cnx, event); |
537 | break; | |
538 | default: | |
61a3c696 ME |
539 | veth_error("Unknown ack type %d from LPAR %d.\n", |
540 | event->base_event.xSubtype, rlp); | |
6403eab1 | 541 | } |
1da177e4 LT |
542 | } |
543 | ||
59f17aeb | 544 | static void veth_handle_int(struct veth_lpevent *event) |
1da177e4 LT |
545 | { |
546 | HvLpIndex rlp = event->base_event.xSourceLp; | |
547 | struct veth_lpar_connection *cnx = veth_cnx[rlp]; | |
548 | unsigned long flags; | |
24562ffa | 549 | int i, acked = 0; |
1da177e4 LT |
550 | |
551 | BUG_ON(! cnx); | |
552 | ||
553 | switch (event->base_event.xSubtype) { | |
59f17aeb | 554 | case VETH_EVENT_CAP: |
1da177e4 LT |
555 | veth_take_cap(cnx, event); |
556 | break; | |
59f17aeb | 557 | case VETH_EVENT_MONITOR: |
1da177e4 LT |
558 | /* do nothing... this'll hang out here til we're dead, |
559 | * and the hypervisor will return it for us. */ | |
560 | break; | |
59f17aeb | 561 | case VETH_EVENT_FRAMES_ACK: |
1da177e4 | 562 | spin_lock_irqsave(&cnx->lock, flags); |
24562ffa | 563 | |
1da177e4 LT |
564 | for (i = 0; i < VETH_MAX_ACKS_PER_MSG; ++i) { |
565 | u16 msgnum = event->u.frames_ack_data.token[i]; | |
566 | ||
24562ffa | 567 | if (msgnum < VETH_NUMBUFFERS) { |
1da177e4 | 568 | veth_recycle_msg(cnx, cnx->msgs + msgnum); |
24562ffa ME |
569 | cnx->outstanding_tx--; |
570 | acked++; | |
571 | } | |
1da177e4 | 572 | } |
24562ffa | 573 | |
e0808494 | 574 | if (acked > 0) { |
24562ffa | 575 | cnx->last_contact = jiffies; |
e0808494 ME |
576 | veth_wake_queues(cnx); |
577 | } | |
24562ffa | 578 | |
1da177e4 | 579 | spin_unlock_irqrestore(&cnx->lock, flags); |
1da177e4 | 580 | break; |
59f17aeb | 581 | case VETH_EVENT_FRAMES: |
1da177e4 LT |
582 | veth_receive(cnx, event); |
583 | break; | |
584 | default: | |
61a3c696 ME |
585 | veth_error("Unknown interrupt type %d from LPAR %d.\n", |
586 | event->base_event.xSubtype, rlp); | |
6403eab1 | 587 | } |
1da177e4 LT |
588 | } |
589 | ||
7d12e780 | 590 | static void veth_handle_event(struct HvLpEvent *event) |
1da177e4 | 591 | { |
59f17aeb | 592 | struct veth_lpevent *veth_event = (struct veth_lpevent *)event; |
1da177e4 | 593 | |
677f8c0d | 594 | if (hvlpevent_is_ack(event)) |
1da177e4 | 595 | veth_handle_ack(veth_event); |
677f8c0d | 596 | else |
1da177e4 LT |
597 | veth_handle_int(veth_event); |
598 | } | |
599 | ||
600 | static int veth_process_caps(struct veth_lpar_connection *cnx) | |
601 | { | |
59f17aeb | 602 | struct veth_cap_data *remote_caps = &cnx->remote_caps; |
1da177e4 LT |
603 | int num_acks_needed; |
604 | ||
605 | /* Convert timer to jiffies */ | |
606 | cnx->ack_timeout = remote_caps->ack_timeout * HZ / 1000000; | |
607 | ||
8e95a202 JP |
608 | if ( (remote_caps->num_buffers == 0) || |
609 | (remote_caps->ack_threshold > VETH_MAX_ACKS_PER_MSG) || | |
610 | (remote_caps->ack_threshold == 0) || | |
611 | (cnx->ack_timeout == 0) ) { | |
61a3c696 ME |
612 | veth_error("Received incompatible capabilities from LPAR %d.\n", |
613 | cnx->remote_lp); | |
1da177e4 LT |
614 | return HvLpEvent_Rc_InvalidSubtypeData; |
615 | } | |
616 | ||
617 | num_acks_needed = (remote_caps->num_buffers | |
618 | / remote_caps->ack_threshold) + 1; | |
619 | ||
620 | /* FIXME: locking on num_ack_events? */ | |
621 | if (cnx->num_ack_events < num_acks_needed) { | |
622 | int num; | |
623 | ||
624 | num = veth_allocate_events(cnx->remote_lp, | |
625 | num_acks_needed-cnx->num_ack_events); | |
626 | if (num > 0) | |
627 | cnx->num_ack_events += num; | |
628 | ||
629 | if (cnx->num_ack_events < num_acks_needed) { | |
61a3c696 ME |
630 | veth_error("Couldn't allocate enough ack events " |
631 | "for LPAR %d.\n", cnx->remote_lp); | |
1da177e4 LT |
632 | |
633 | return HvLpEvent_Rc_BufferNotAvailable; | |
634 | } | |
635 | } | |
636 | ||
637 | ||
638 | return HvLpEvent_Rc_Good; | |
639 | } | |
640 | ||
641 | /* FIXME: The gotos here are a bit dubious */ | |
c4028958 | 642 | static void veth_statemachine(struct work_struct *work) |
1da177e4 | 643 | { |
c4028958 DH |
644 | struct veth_lpar_connection *cnx = |
645 | container_of(work, struct veth_lpar_connection, | |
646 | statemachine_wq.work); | |
1da177e4 LT |
647 | int rlp = cnx->remote_lp; |
648 | int rc; | |
649 | ||
650 | spin_lock_irq(&cnx->lock); | |
651 | ||
652 | restart: | |
653 | if (cnx->state & VETH_STATE_RESET) { | |
1da177e4 LT |
654 | if (cnx->state & VETH_STATE_OPEN) |
655 | HvCallEvent_closeLpEventPath(cnx->remote_lp, | |
656 | HvLpEvent_Type_VirtualLan); | |
657 | ||
abfda471 ME |
658 | /* |
659 | * Reset ack data. This prevents the ack_timer actually | |
660 | * doing anything, even if it runs one more time when | |
661 | * we drop the lock below. | |
662 | */ | |
1da177e4 LT |
663 | memset(&cnx->pending_acks, 0xff, sizeof (cnx->pending_acks)); |
664 | cnx->num_pending_acks = 0; | |
665 | ||
666 | cnx->state &= ~(VETH_STATE_RESET | VETH_STATE_SENTMON | |
667 | | VETH_STATE_OPEN | VETH_STATE_SENTCAPS | |
668 | | VETH_STATE_GOTCAPACK | VETH_STATE_GOTCAPS | |
669 | | VETH_STATE_SENTCAPACK | VETH_STATE_READY); | |
670 | ||
671 | /* Clean up any leftover messages */ | |
24562ffa ME |
672 | if (cnx->msgs) { |
673 | int i; | |
1da177e4 LT |
674 | for (i = 0; i < VETH_NUMBUFFERS; ++i) |
675 | veth_recycle_msg(cnx, cnx->msgs + i); | |
24562ffa | 676 | } |
e0808494 | 677 | |
24562ffa | 678 | cnx->outstanding_tx = 0; |
e0808494 | 679 | veth_wake_queues(cnx); |
abfda471 ME |
680 | |
681 | /* Drop the lock so we can do stuff that might sleep or | |
682 | * take other locks. */ | |
1da177e4 | 683 | spin_unlock_irq(&cnx->lock); |
abfda471 ME |
684 | |
685 | del_timer_sync(&cnx->ack_timer); | |
24562ffa ME |
686 | del_timer_sync(&cnx->reset_timer); |
687 | ||
1da177e4 | 688 | spin_lock_irq(&cnx->lock); |
abfda471 | 689 | |
1da177e4 LT |
690 | if (cnx->state & VETH_STATE_RESET) |
691 | goto restart; | |
58c5900b ME |
692 | |
693 | /* Hack, wait for the other end to reset itself. */ | |
694 | if (! (cnx->state & VETH_STATE_SHUTDOWN)) { | |
695 | schedule_delayed_work(&cnx->statemachine_wq, 5 * HZ); | |
696 | goto out; | |
697 | } | |
1da177e4 LT |
698 | } |
699 | ||
700 | if (cnx->state & VETH_STATE_SHUTDOWN) | |
701 | /* It's all over, do nothing */ | |
702 | goto out; | |
703 | ||
704 | if ( !(cnx->state & VETH_STATE_OPEN) ) { | |
705 | if (! cnx->msgs || (cnx->num_events < (2 + VETH_NUMBUFFERS)) ) | |
706 | goto cant_cope; | |
707 | ||
708 | HvCallEvent_openLpEventPath(rlp, HvLpEvent_Type_VirtualLan); | |
709 | cnx->src_inst = | |
710 | HvCallEvent_getSourceLpInstanceId(rlp, | |
711 | HvLpEvent_Type_VirtualLan); | |
712 | cnx->dst_inst = | |
713 | HvCallEvent_getTargetLpInstanceId(rlp, | |
714 | HvLpEvent_Type_VirtualLan); | |
715 | cnx->state |= VETH_STATE_OPEN; | |
716 | } | |
717 | ||
8e95a202 JP |
718 | if ( (cnx->state & VETH_STATE_OPEN) && |
719 | !(cnx->state & VETH_STATE_SENTMON) ) { | |
59f17aeb | 720 | rc = veth_signalevent(cnx, VETH_EVENT_MONITOR, |
1da177e4 LT |
721 | HvLpEvent_AckInd_DoAck, |
722 | HvLpEvent_AckType_DeferredAck, | |
723 | 0, 0, 0, 0, 0, 0); | |
724 | ||
725 | if (rc == HvLpEvent_Rc_Good) { | |
726 | cnx->state |= VETH_STATE_SENTMON; | |
727 | } else { | |
8e95a202 JP |
728 | if ( (rc != HvLpEvent_Rc_PartitionDead) && |
729 | (rc != HvLpEvent_Rc_PathClosed) ) | |
61a3c696 ME |
730 | veth_error("Error sending monitor to LPAR %d, " |
731 | "rc = %d\n", rlp, rc); | |
1da177e4 LT |
732 | |
733 | /* Oh well, hope we get a cap from the other | |
734 | * end and do better when that kicks us */ | |
735 | goto out; | |
736 | } | |
737 | } | |
738 | ||
8e95a202 JP |
739 | if ( (cnx->state & VETH_STATE_OPEN) && |
740 | !(cnx->state & VETH_STATE_SENTCAPS)) { | |
1da177e4 LT |
741 | u64 *rawcap = (u64 *)&cnx->local_caps; |
742 | ||
59f17aeb | 743 | rc = veth_signalevent(cnx, VETH_EVENT_CAP, |
1da177e4 LT |
744 | HvLpEvent_AckInd_DoAck, |
745 | HvLpEvent_AckType_ImmediateAck, | |
746 | 0, rawcap[0], rawcap[1], rawcap[2], | |
747 | rawcap[3], rawcap[4]); | |
748 | ||
749 | if (rc == HvLpEvent_Rc_Good) { | |
750 | cnx->state |= VETH_STATE_SENTCAPS; | |
751 | } else { | |
8e95a202 JP |
752 | if ( (rc != HvLpEvent_Rc_PartitionDead) && |
753 | (rc != HvLpEvent_Rc_PathClosed) ) | |
61a3c696 ME |
754 | veth_error("Error sending caps to LPAR %d, " |
755 | "rc = %d\n", rlp, rc); | |
756 | ||
1da177e4 LT |
757 | /* Oh well, hope we get a cap from the other |
758 | * end and do better when that kicks us */ | |
759 | goto out; | |
760 | } | |
761 | } | |
762 | ||
8e95a202 JP |
763 | if ((cnx->state & VETH_STATE_GOTCAPS) && |
764 | !(cnx->state & VETH_STATE_SENTCAPACK)) { | |
59f17aeb | 765 | struct veth_cap_data *remote_caps = &cnx->remote_caps; |
1da177e4 LT |
766 | |
767 | memcpy(remote_caps, &cnx->cap_event.u.caps_data, | |
768 | sizeof(*remote_caps)); | |
769 | ||
770 | spin_unlock_irq(&cnx->lock); | |
771 | rc = veth_process_caps(cnx); | |
772 | spin_lock_irq(&cnx->lock); | |
773 | ||
774 | /* We dropped the lock, so recheck for anything which | |
775 | * might mess us up */ | |
776 | if (cnx->state & (VETH_STATE_RESET|VETH_STATE_SHUTDOWN)) | |
777 | goto restart; | |
778 | ||
779 | cnx->cap_event.base_event.xRc = rc; | |
780 | HvCallEvent_ackLpEvent((struct HvLpEvent *)&cnx->cap_event); | |
781 | if (rc == HvLpEvent_Rc_Good) | |
782 | cnx->state |= VETH_STATE_SENTCAPACK; | |
783 | else | |
784 | goto cant_cope; | |
785 | } | |
786 | ||
8e95a202 JP |
787 | if ((cnx->state & VETH_STATE_GOTCAPACK) && |
788 | (cnx->state & VETH_STATE_GOTCAPS) && | |
789 | !(cnx->state & VETH_STATE_READY)) { | |
1da177e4 LT |
790 | if (cnx->cap_ack_event.base_event.xRc == HvLpEvent_Rc_Good) { |
791 | /* Start the ACK timer */ | |
792 | cnx->ack_timer.expires = jiffies + cnx->ack_timeout; | |
793 | add_timer(&cnx->ack_timer); | |
794 | cnx->state |= VETH_STATE_READY; | |
795 | } else { | |
61a3c696 ME |
796 | veth_error("Caps rejected by LPAR %d, rc = %d\n", |
797 | rlp, cnx->cap_ack_event.base_event.xRc); | |
1da177e4 LT |
798 | goto cant_cope; |
799 | } | |
800 | } | |
801 | ||
802 | out: | |
803 | spin_unlock_irq(&cnx->lock); | |
804 | return; | |
805 | ||
806 | cant_cope: | |
807 | /* FIXME: we get here if something happens we really can't | |
808 | * cope with. The link will never work once we get here, and | |
809 | * all we can do is not lock the rest of the system up */ | |
61a3c696 ME |
810 | veth_error("Unrecoverable error on connection to LPAR %d, shutting down" |
811 | " (state = 0x%04lx)\n", rlp, cnx->state); | |
1da177e4 LT |
812 | cnx->state |= VETH_STATE_SHUTDOWN; |
813 | spin_unlock_irq(&cnx->lock); | |
814 | } | |
815 | ||
816 | static int veth_init_connection(u8 rlp) | |
817 | { | |
818 | struct veth_lpar_connection *cnx; | |
819 | struct veth_msg *msgs; | |
78be3d2f | 820 | int i; |
1da177e4 | 821 | |
8e95a202 JP |
822 | if ( (rlp == this_lp) || |
823 | ! HvLpConfig_doLpsCommunicateOnVirtualLan(this_lp, rlp) ) | |
1da177e4 LT |
824 | return 0; |
825 | ||
dd00cc48 | 826 | cnx = kzalloc(sizeof(*cnx), GFP_KERNEL); |
1da177e4 LT |
827 | if (! cnx) |
828 | return -ENOMEM; | |
1da177e4 LT |
829 | |
830 | cnx->remote_lp = rlp; | |
831 | spin_lock_init(&cnx->lock); | |
c4028958 | 832 | INIT_DELAYED_WORK(&cnx->statemachine_wq, veth_statemachine); |
24562ffa | 833 | |
1da177e4 LT |
834 | init_timer(&cnx->ack_timer); |
835 | cnx->ack_timer.function = veth_timed_ack; | |
836 | cnx->ack_timer.data = (unsigned long) cnx; | |
24562ffa ME |
837 | |
838 | init_timer(&cnx->reset_timer); | |
839 | cnx->reset_timer.function = veth_timed_reset; | |
840 | cnx->reset_timer.data = (unsigned long) cnx; | |
841 | cnx->reset_timeout = 5 * HZ * (VETH_ACKTIMEOUT / 1000000); | |
842 | ||
1da177e4 LT |
843 | memset(&cnx->pending_acks, 0xff, sizeof (cnx->pending_acks)); |
844 | ||
845 | veth_cnx[rlp] = cnx; | |
846 | ||
f0c129ca ME |
847 | /* This gets us 1 reference, which is held on behalf of the driver |
848 | * infrastructure. It's released at module unload. */ | |
f9cb074b | 849 | kobject_init(&cnx->kobject, &veth_lpar_connection_ktype); |
f0c129ca | 850 | |
dd00cc48 | 851 | msgs = kcalloc(VETH_NUMBUFFERS, sizeof(struct veth_msg), GFP_KERNEL); |
1da177e4 | 852 | if (! msgs) { |
61a3c696 | 853 | veth_error("Can't allocate buffers for LPAR %d.\n", rlp); |
1da177e4 LT |
854 | return -ENOMEM; |
855 | } | |
856 | ||
857 | cnx->msgs = msgs; | |
1da177e4 LT |
858 | |
859 | for (i = 0; i < VETH_NUMBUFFERS; i++) { | |
860 | msgs[i].token = i; | |
861 | veth_stack_push(cnx, msgs + i); | |
862 | } | |
863 | ||
864 | cnx->num_events = veth_allocate_events(rlp, 2 + VETH_NUMBUFFERS); | |
865 | ||
866 | if (cnx->num_events < (2 + VETH_NUMBUFFERS)) { | |
61a3c696 | 867 | veth_error("Can't allocate enough events for LPAR %d.\n", rlp); |
1da177e4 LT |
868 | return -ENOMEM; |
869 | } | |
870 | ||
871 | cnx->local_caps.num_buffers = VETH_NUMBUFFERS; | |
872 | cnx->local_caps.ack_threshold = ACK_THRESHOLD; | |
873 | cnx->local_caps.ack_timeout = VETH_ACKTIMEOUT; | |
874 | ||
875 | return 0; | |
876 | } | |
877 | ||
f0c129ca | 878 | static void veth_stop_connection(struct veth_lpar_connection *cnx) |
1da177e4 | 879 | { |
f0c129ca | 880 | if (!cnx) |
1da177e4 LT |
881 | return; |
882 | ||
883 | spin_lock_irq(&cnx->lock); | |
884 | cnx->state |= VETH_STATE_RESET | VETH_STATE_SHUTDOWN; | |
885 | veth_kick_statemachine(cnx); | |
886 | spin_unlock_irq(&cnx->lock); | |
887 | ||
9beb4896 TH |
888 | /* ensure the statemachine runs now and waits for its completion */ |
889 | flush_delayed_work_sync(&cnx->statemachine_wq); | |
ec60beeb ME |
890 | } |
891 | ||
f0c129ca | 892 | static void veth_destroy_connection(struct veth_lpar_connection *cnx) |
ec60beeb | 893 | { |
f0c129ca | 894 | if (!cnx) |
ec60beeb | 895 | return; |
1da177e4 | 896 | |
1da177e4 LT |
897 | if (cnx->num_events > 0) |
898 | mf_deallocate_lp_events(cnx->remote_lp, | |
899 | HvLpEvent_Type_VirtualLan, | |
900 | cnx->num_events, | |
901 | NULL, NULL); | |
902 | if (cnx->num_ack_events > 0) | |
903 | mf_deallocate_lp_events(cnx->remote_lp, | |
904 | HvLpEvent_Type_VirtualLan, | |
905 | cnx->num_ack_events, | |
906 | NULL, NULL); | |
1da177e4 LT |
907 | |
908 | kfree(cnx->msgs); | |
f0c129ca | 909 | veth_cnx[cnx->remote_lp] = NULL; |
1da177e4 | 910 | kfree(cnx); |
f0c129ca ME |
911 | } |
912 | ||
913 | static void veth_release_connection(struct kobject *kobj) | |
914 | { | |
915 | struct veth_lpar_connection *cnx; | |
916 | cnx = container_of(kobj, struct veth_lpar_connection, kobject); | |
917 | veth_stop_connection(cnx); | |
918 | veth_destroy_connection(cnx); | |
1da177e4 LT |
919 | } |
920 | ||
921 | /* | |
922 | * net_device code | |
923 | */ | |
924 | ||
925 | static int veth_open(struct net_device *dev) | |
926 | { | |
1da177e4 LT |
927 | netif_start_queue(dev); |
928 | return 0; | |
929 | } | |
930 | ||
931 | static int veth_close(struct net_device *dev) | |
932 | { | |
933 | netif_stop_queue(dev); | |
934 | return 0; | |
935 | } | |
936 | ||
1da177e4 LT |
937 | static int veth_change_mtu(struct net_device *dev, int new_mtu) |
938 | { | |
939 | if ((new_mtu < 68) || (new_mtu > VETH_MAX_MTU)) | |
940 | return -EINVAL; | |
941 | dev->mtu = new_mtu; | |
942 | return 0; | |
943 | } | |
944 | ||
945 | static void veth_set_multicast_list(struct net_device *dev) | |
946 | { | |
454d7c9b | 947 | struct veth_port *port = netdev_priv(dev); |
1da177e4 LT |
948 | unsigned long flags; |
949 | ||
950 | write_lock_irqsave(&port->mcast_gate, flags); | |
951 | ||
2a5391a1 | 952 | if ((dev->flags & IFF_PROMISC) || (dev->flags & IFF_ALLMULTI) || |
4cd24eaf | 953 | (netdev_mc_count(dev) > VETH_MAX_MCAST)) { |
1da177e4 | 954 | port->promiscuous = 1; |
1da177e4 | 955 | } else { |
22bedad3 | 956 | struct netdev_hw_addr *ha; |
1da177e4 | 957 | |
2a5391a1 ME |
958 | port->promiscuous = 0; |
959 | ||
1da177e4 LT |
960 | /* Update table */ |
961 | port->num_mcast = 0; | |
962 | ||
22bedad3 JP |
963 | netdev_for_each_mc_addr(ha, dev) { |
964 | u8 *addr = ha->addr; | |
1da177e4 LT |
965 | u64 xaddr = 0; |
966 | ||
eb31c4e0 TK |
967 | memcpy(&xaddr, addr, ETH_ALEN); |
968 | port->mcast_addr[port->num_mcast] = xaddr; | |
969 | port->num_mcast++; | |
1da177e4 LT |
970 | } |
971 | } | |
972 | ||
973 | write_unlock_irqrestore(&port->mcast_gate, flags); | |
974 | } | |
975 | ||
976 | static void veth_get_drvinfo(struct net_device *dev, struct ethtool_drvinfo *info) | |
977 | { | |
ee05f031 | 978 | strncpy(info->driver, DRV_NAME, sizeof(info->driver) - 1); |
1da177e4 | 979 | info->driver[sizeof(info->driver) - 1] = '\0'; |
ee05f031 ME |
980 | strncpy(info->version, DRV_VERSION, sizeof(info->version) - 1); |
981 | info->version[sizeof(info->version) - 1] = '\0'; | |
1da177e4 LT |
982 | } |
983 | ||
984 | static int veth_get_settings(struct net_device *dev, struct ethtool_cmd *ecmd) | |
985 | { | |
986 | ecmd->supported = (SUPPORTED_1000baseT_Full | |
987 | | SUPPORTED_Autoneg | SUPPORTED_FIBRE); | |
988 | ecmd->advertising = (SUPPORTED_1000baseT_Full | |
989 | | SUPPORTED_Autoneg | SUPPORTED_FIBRE); | |
990 | ecmd->port = PORT_FIBRE; | |
991 | ecmd->transceiver = XCVR_INTERNAL; | |
992 | ecmd->phy_address = 0; | |
993 | ecmd->speed = SPEED_1000; | |
994 | ecmd->duplex = DUPLEX_FULL; | |
995 | ecmd->autoneg = AUTONEG_ENABLE; | |
996 | ecmd->maxtxpkt = 120; | |
997 | ecmd->maxrxpkt = 120; | |
998 | return 0; | |
999 | } | |
1000 | ||
7282d491 | 1001 | static const struct ethtool_ops ops = { |
1da177e4 LT |
1002 | .get_drvinfo = veth_get_drvinfo, |
1003 | .get_settings = veth_get_settings, | |
ed4ba4b5 | 1004 | .get_link = ethtool_op_get_link, |
1da177e4 LT |
1005 | }; |
1006 | ||
c2775360 AB |
1007 | static const struct net_device_ops veth_netdev_ops = { |
1008 | .ndo_open = veth_open, | |
1009 | .ndo_stop = veth_close, | |
1010 | .ndo_start_xmit = veth_start_xmit, | |
1011 | .ndo_change_mtu = veth_change_mtu, | |
afc4b13d | 1012 | .ndo_set_rx_mode = veth_set_multicast_list, |
c2775360 AB |
1013 | .ndo_set_mac_address = NULL, |
1014 | .ndo_validate_addr = eth_validate_addr, | |
1015 | }; | |
1016 | ||
57f78ab3 | 1017 | static struct net_device *veth_probe_one(int vlan, |
07fb3f45 | 1018 | struct vio_dev *vio_dev) |
1da177e4 LT |
1019 | { |
1020 | struct net_device *dev; | |
1021 | struct veth_port *port; | |
07fb3f45 | 1022 | struct device *vdev = &vio_dev->dev; |
1da177e4 | 1023 | int i, rc; |
07fb3f45 SR |
1024 | const unsigned char *mac_addr; |
1025 | ||
1026 | mac_addr = vio_get_attribute(vio_dev, "local-mac-address", NULL); | |
1027 | if (mac_addr == NULL) | |
1028 | mac_addr = vio_get_attribute(vio_dev, "mac-address", NULL); | |
1029 | if (mac_addr == NULL) { | |
1030 | veth_error("Unable to fetch MAC address from device tree.\n"); | |
1031 | return NULL; | |
1032 | } | |
1da177e4 LT |
1033 | |
1034 | dev = alloc_etherdev(sizeof (struct veth_port)); | |
41de8d4c | 1035 | if (!dev) |
1da177e4 | 1036 | return NULL; |
1da177e4 | 1037 | |
454d7c9b | 1038 | port = netdev_priv(dev); |
1da177e4 | 1039 | |
e0808494 | 1040 | spin_lock_init(&port->queue_lock); |
1da177e4 | 1041 | rwlock_init(&port->mcast_gate); |
e0808494 | 1042 | port->stopped_map = 0; |
1da177e4 LT |
1043 | |
1044 | for (i = 0; i < HVMAXARCHITECTEDLPS; i++) { | |
1045 | HvLpVirtualLanIndexMap map; | |
1046 | ||
1047 | if (i == this_lp) | |
1048 | continue; | |
1049 | map = HvLpConfig_getVirtualLanIndexMapForLp(i); | |
1050 | if (map & (0x8000 >> vlan)) | |
1051 | port->lpar_map |= (1 << i); | |
1052 | } | |
1053 | port->dev = vdev; | |
1054 | ||
07fb3f45 | 1055 | memcpy(dev->dev_addr, mac_addr, ETH_ALEN); |
1da177e4 LT |
1056 | |
1057 | dev->mtu = VETH_MAX_MTU; | |
1058 | ||
07fb3f45 | 1059 | memcpy(&port->mac_addr, mac_addr, ETH_ALEN); |
1da177e4 | 1060 | |
3ef8e4e9 | 1061 | dev->netdev_ops = &veth_netdev_ops; |
1da177e4 LT |
1062 | SET_ETHTOOL_OPS(dev, &ops); |
1063 | ||
1da177e4 LT |
1064 | SET_NETDEV_DEV(dev, vdev); |
1065 | ||
1066 | rc = register_netdev(dev); | |
1067 | if (rc != 0) { | |
61a3c696 | 1068 | veth_error("Failed registering net device for vlan%d.\n", vlan); |
1da177e4 LT |
1069 | free_netdev(dev); |
1070 | return NULL; | |
1071 | } | |
1072 | ||
f9cb074b | 1073 | kobject_init(&port->kobject, &veth_port_ktype); |
b2d6db58 | 1074 | if (0 != kobject_add(&port->kobject, &dev->dev.kobj, "veth_port")) |
07a5c172 ME |
1075 | veth_error("Failed adding port for %s to sysfs.\n", dev->name); |
1076 | ||
61a3c696 ME |
1077 | veth_info("%s attached to iSeries vlan %d (LPAR map = 0x%.4X)\n", |
1078 | dev->name, vlan, port->lpar_map); | |
1da177e4 LT |
1079 | |
1080 | return dev; | |
1081 | } | |
1082 | ||
1083 | /* | |
1084 | * Tx path | |
1085 | */ | |
1086 | ||
1087 | static int veth_transmit_to_one(struct sk_buff *skb, HvLpIndex rlp, | |
1088 | struct net_device *dev) | |
1089 | { | |
1090 | struct veth_lpar_connection *cnx = veth_cnx[rlp]; | |
454d7c9b | 1091 | struct veth_port *port = netdev_priv(dev); |
1da177e4 | 1092 | HvLpEvent_Rc rc; |
1da177e4 | 1093 | struct veth_msg *msg = NULL; |
1da177e4 LT |
1094 | unsigned long flags; |
1095 | ||
db5e8718 | 1096 | if (! cnx) |
1da177e4 | 1097 | return 0; |
1da177e4 LT |
1098 | |
1099 | spin_lock_irqsave(&cnx->lock, flags); | |
1100 | ||
f27eff1f | 1101 | if (! (cnx->state & VETH_STATE_READY)) |
db5e8718 | 1102 | goto no_error; |
1da177e4 | 1103 | |
db5e8718 | 1104 | if ((skb->len - ETH_HLEN) > VETH_MAX_MTU) |
1da177e4 LT |
1105 | goto drop; |
1106 | ||
1107 | msg = veth_stack_pop(cnx); | |
db5e8718 | 1108 | if (! msg) |
1da177e4 | 1109 | goto drop; |
1da177e4 | 1110 | |
b08bd5c0 | 1111 | msg->in_use = 1; |
db5e8718 | 1112 | msg->skb = skb_get(skb); |
b08bd5c0 | 1113 | |
cbf9074c ME |
1114 | msg->data.addr[0] = dma_map_single(port->dev, skb->data, |
1115 | skb->len, DMA_TO_DEVICE); | |
1da177e4 | 1116 | |
8d8bb39b | 1117 | if (dma_mapping_error(port->dev, msg->data.addr[0])) |
1da177e4 LT |
1118 | goto recycle_and_drop; |
1119 | ||
1da177e4 | 1120 | msg->dev = port->dev; |
cbf9074c | 1121 | msg->data.len[0] = skb->len; |
1da177e4 | 1122 | msg->data.eofmask = 1 << VETH_EOF_SHIFT; |
cbf9074c | 1123 | |
59f17aeb | 1124 | rc = veth_signaldata(cnx, VETH_EVENT_FRAMES, msg->token, &msg->data); |
1da177e4 LT |
1125 | |
1126 | if (rc != HvLpEvent_Rc_Good) | |
1127 | goto recycle_and_drop; | |
1128 | ||
24562ffa ME |
1129 | /* If the timer's not already running, start it now. */ |
1130 | if (0 == cnx->outstanding_tx) | |
1131 | mod_timer(&cnx->reset_timer, jiffies + cnx->reset_timeout); | |
1132 | ||
1133 | cnx->last_contact = jiffies; | |
1134 | cnx->outstanding_tx++; | |
1135 | ||
e0808494 ME |
1136 | if (veth_stack_is_empty(cnx)) |
1137 | veth_stop_queues(cnx); | |
1138 | ||
db5e8718 | 1139 | no_error: |
1da177e4 LT |
1140 | spin_unlock_irqrestore(&cnx->lock, flags); |
1141 | return 0; | |
1142 | ||
1143 | recycle_and_drop: | |
1da177e4 LT |
1144 | veth_recycle_msg(cnx, msg); |
1145 | drop: | |
1da177e4 | 1146 | spin_unlock_irqrestore(&cnx->lock, flags); |
db5e8718 | 1147 | return 1; |
1da177e4 LT |
1148 | } |
1149 | ||
db5e8718 | 1150 | static void veth_transmit_to_many(struct sk_buff *skb, |
1da177e4 LT |
1151 | HvLpIndexMap lpmask, |
1152 | struct net_device *dev) | |
1153 | { | |
db5e8718 ME |
1154 | int i, success, error; |
1155 | ||
1156 | success = error = 0; | |
1da177e4 LT |
1157 | |
1158 | for (i = 0; i < HVMAXARCHITECTEDLPS; i++) { | |
1159 | if ((lpmask & (1 << i)) == 0) | |
1160 | continue; | |
1161 | ||
db5e8718 ME |
1162 | if (veth_transmit_to_one(skb, i, dev)) |
1163 | error = 1; | |
1164 | else | |
1165 | success = 1; | |
1da177e4 LT |
1166 | } |
1167 | ||
db5e8718 | 1168 | if (error) |
09f75cd7 | 1169 | dev->stats.tx_errors++; |
db5e8718 ME |
1170 | |
1171 | if (success) { | |
09f75cd7 JG |
1172 | dev->stats.tx_packets++; |
1173 | dev->stats.tx_bytes += skb->len; | |
1da177e4 | 1174 | } |
1da177e4 LT |
1175 | } |
1176 | ||
1177 | static int veth_start_xmit(struct sk_buff *skb, struct net_device *dev) | |
1178 | { | |
1179 | unsigned char *frame = skb->data; | |
454d7c9b | 1180 | struct veth_port *port = netdev_priv(dev); |
1da177e4 LT |
1181 | HvLpIndexMap lpmask; |
1182 | ||
8bc3e1d8 | 1183 | if (is_unicast_ether_addr(frame)) { |
1da177e4 LT |
1184 | /* unicast packet */ |
1185 | HvLpIndex rlp = frame[5]; | |
1186 | ||
1187 | if ( ! ((1 << rlp) & port->lpar_map) ) { | |
1188 | dev_kfree_skb(skb); | |
6ed10654 | 1189 | return NETDEV_TX_OK; |
1da177e4 LT |
1190 | } |
1191 | ||
1192 | lpmask = 1 << rlp; | |
1193 | } else { | |
1194 | lpmask = port->lpar_map; | |
1195 | } | |
1196 | ||
e0808494 | 1197 | veth_transmit_to_many(skb, lpmask, dev); |
1da177e4 | 1198 | |
e0808494 | 1199 | dev_kfree_skb(skb); |
1da177e4 | 1200 | |
6ed10654 | 1201 | return NETDEV_TX_OK; |
1da177e4 LT |
1202 | } |
1203 | ||
b08bd5c0 | 1204 | /* You must hold the connection's lock when you call this function. */ |
1da177e4 LT |
1205 | static void veth_recycle_msg(struct veth_lpar_connection *cnx, |
1206 | struct veth_msg *msg) | |
1207 | { | |
1208 | u32 dma_address, dma_length; | |
1209 | ||
b08bd5c0 ME |
1210 | if (msg->in_use) { |
1211 | msg->in_use = 0; | |
1da177e4 LT |
1212 | dma_address = msg->data.addr[0]; |
1213 | dma_length = msg->data.len[0]; | |
1214 | ||
8d8bb39b | 1215 | if (!dma_mapping_error(msg->dev, dma_address)) |
cbf9074c ME |
1216 | dma_unmap_single(msg->dev, dma_address, dma_length, |
1217 | DMA_TO_DEVICE); | |
1da177e4 LT |
1218 | |
1219 | if (msg->skb) { | |
1220 | dev_kfree_skb_any(msg->skb); | |
1221 | msg->skb = NULL; | |
1222 | } | |
1223 | ||
1224 | memset(&msg->data, 0, sizeof(msg->data)); | |
1225 | veth_stack_push(cnx, msg); | |
61a3c696 ME |
1226 | } else if (cnx->state & VETH_STATE_OPEN) { |
1227 | veth_error("Non-pending frame (# %d) acked by LPAR %d.\n", | |
1228 | cnx->remote_lp, msg->token); | |
1229 | } | |
1da177e4 LT |
1230 | } |
1231 | ||
e0808494 | 1232 | static void veth_wake_queues(struct veth_lpar_connection *cnx) |
1da177e4 LT |
1233 | { |
1234 | int i; | |
e0808494 | 1235 | |
1da177e4 LT |
1236 | for (i = 0; i < HVMAXARCHITECTEDVIRTUALLANS; i++) { |
1237 | struct net_device *dev = veth_dev[i]; | |
1238 | struct veth_port *port; | |
1239 | unsigned long flags; | |
1240 | ||
1241 | if (! dev) | |
1242 | continue; | |
1243 | ||
454d7c9b | 1244 | port = netdev_priv(dev); |
1da177e4 LT |
1245 | |
1246 | if (! (port->lpar_map & (1<<cnx->remote_lp))) | |
1247 | continue; | |
1248 | ||
e0808494 ME |
1249 | spin_lock_irqsave(&port->queue_lock, flags); |
1250 | ||
1251 | port->stopped_map &= ~(1 << cnx->remote_lp); | |
1252 | ||
1253 | if (0 == port->stopped_map && netif_queue_stopped(dev)) { | |
1254 | veth_debug("cnx %d: woke queue for %s.\n", | |
1255 | cnx->remote_lp, dev->name); | |
1256 | netif_wake_queue(dev); | |
1da177e4 | 1257 | } |
e0808494 ME |
1258 | spin_unlock_irqrestore(&port->queue_lock, flags); |
1259 | } | |
1260 | } | |
1261 | ||
1262 | static void veth_stop_queues(struct veth_lpar_connection *cnx) | |
1263 | { | |
1264 | int i; | |
1265 | ||
1266 | for (i = 0; i < HVMAXARCHITECTEDVIRTUALLANS; i++) { | |
1267 | struct net_device *dev = veth_dev[i]; | |
1268 | struct veth_port *port; | |
1269 | ||
1270 | if (! dev) | |
1271 | continue; | |
1272 | ||
454d7c9b | 1273 | port = netdev_priv(dev); |
e0808494 ME |
1274 | |
1275 | /* If this cnx is not on the vlan for this port, continue */ | |
1276 | if (! (port->lpar_map & (1 << cnx->remote_lp))) | |
1277 | continue; | |
1278 | ||
1279 | spin_lock(&port->queue_lock); | |
1280 | ||
1281 | netif_stop_queue(dev); | |
1282 | port->stopped_map |= (1 << cnx->remote_lp); | |
1283 | ||
1284 | veth_debug("cnx %d: stopped queue for %s, map = 0x%x.\n", | |
1285 | cnx->remote_lp, dev->name, port->stopped_map); | |
1286 | ||
1287 | spin_unlock(&port->queue_lock); | |
1da177e4 LT |
1288 | } |
1289 | } | |
1290 | ||
24562ffa ME |
1291 | static void veth_timed_reset(unsigned long ptr) |
1292 | { | |
1293 | struct veth_lpar_connection *cnx = (struct veth_lpar_connection *)ptr; | |
1294 | unsigned long trigger_time, flags; | |
1295 | ||
1296 | /* FIXME is it possible this fires after veth_stop_connection()? | |
1297 | * That would reschedule the statemachine for 5 seconds and probably | |
1298 | * execute it after the module's been unloaded. Hmm. */ | |
1299 | ||
1300 | spin_lock_irqsave(&cnx->lock, flags); | |
1301 | ||
1302 | if (cnx->outstanding_tx > 0) { | |
1303 | trigger_time = cnx->last_contact + cnx->reset_timeout; | |
1304 | ||
1305 | if (trigger_time < jiffies) { | |
1306 | cnx->state |= VETH_STATE_RESET; | |
1307 | veth_kick_statemachine(cnx); | |
1308 | veth_error("%d packets not acked by LPAR %d within %d " | |
1309 | "seconds, resetting.\n", | |
1310 | cnx->outstanding_tx, cnx->remote_lp, | |
1311 | cnx->reset_timeout / HZ); | |
1312 | } else { | |
1313 | /* Reschedule the timer */ | |
1314 | trigger_time = jiffies + cnx->reset_timeout; | |
1315 | mod_timer(&cnx->reset_timer, trigger_time); | |
1316 | } | |
1317 | } | |
1318 | ||
1319 | spin_unlock_irqrestore(&cnx->lock, flags); | |
1320 | } | |
1321 | ||
1da177e4 LT |
1322 | /* |
1323 | * Rx path | |
1324 | */ | |
1325 | ||
1326 | static inline int veth_frame_wanted(struct veth_port *port, u64 mac_addr) | |
1327 | { | |
1328 | int wanted = 0; | |
1329 | int i; | |
1330 | unsigned long flags; | |
1331 | ||
1332 | if ( (mac_addr == port->mac_addr) || (mac_addr == 0xffffffffffff0000) ) | |
1333 | return 1; | |
1334 | ||
1da177e4 LT |
1335 | read_lock_irqsave(&port->mcast_gate, flags); |
1336 | ||
2a5391a1 | 1337 | if (port->promiscuous) { |
1da177e4 LT |
1338 | wanted = 1; |
1339 | goto out; | |
1340 | } | |
1341 | ||
1342 | for (i = 0; i < port->num_mcast; ++i) { | |
1343 | if (port->mcast_addr[i] == mac_addr) { | |
1344 | wanted = 1; | |
1345 | break; | |
1346 | } | |
1347 | } | |
1348 | ||
1349 | out: | |
1350 | read_unlock_irqrestore(&port->mcast_gate, flags); | |
1351 | ||
1352 | return wanted; | |
1353 | } | |
1354 | ||
1355 | struct dma_chunk { | |
1356 | u64 addr; | |
1357 | u64 size; | |
1358 | }; | |
1359 | ||
1360 | #define VETH_MAX_PAGES_PER_FRAME ( (VETH_MAX_MTU+PAGE_SIZE-2)/PAGE_SIZE + 1 ) | |
1361 | ||
1362 | static inline void veth_build_dma_list(struct dma_chunk *list, | |
1363 | unsigned char *p, unsigned long length) | |
1364 | { | |
1365 | unsigned long done; | |
1366 | int i = 1; | |
1367 | ||
af901ca1 | 1368 | /* FIXME: skbs are contiguous in real addresses. Do we |
1da177e4 LT |
1369 | * really need to break it into PAGE_SIZE chunks, or can we do |
1370 | * it just at the granularity of iSeries real->absolute | |
1371 | * mapping? Indeed, given the way the allocator works, can we | |
1372 | * count on them being absolutely contiguous? */ | |
426c1a11 | 1373 | list[0].addr = iseries_hv_addr(p); |
1da177e4 LT |
1374 | list[0].size = min(length, |
1375 | PAGE_SIZE - ((unsigned long)p & ~PAGE_MASK)); | |
1376 | ||
1377 | done = list[0].size; | |
1378 | while (done < length) { | |
426c1a11 | 1379 | list[i].addr = iseries_hv_addr(p + done); |
1da177e4 LT |
1380 | list[i].size = min(length-done, PAGE_SIZE); |
1381 | done += list[i].size; | |
1382 | i++; | |
1383 | } | |
1384 | } | |
1385 | ||
1386 | static void veth_flush_acks(struct veth_lpar_connection *cnx) | |
1387 | { | |
1388 | HvLpEvent_Rc rc; | |
1389 | ||
59f17aeb | 1390 | rc = veth_signaldata(cnx, VETH_EVENT_FRAMES_ACK, |
1da177e4 LT |
1391 | 0, &cnx->pending_acks); |
1392 | ||
1393 | if (rc != HvLpEvent_Rc_Good) | |
61a3c696 ME |
1394 | veth_error("Failed acking frames from LPAR %d, rc = %d\n", |
1395 | cnx->remote_lp, (int)rc); | |
1da177e4 LT |
1396 | |
1397 | cnx->num_pending_acks = 0; | |
1398 | memset(&cnx->pending_acks, 0xff, sizeof(cnx->pending_acks)); | |
1399 | } | |
1400 | ||
1401 | static void veth_receive(struct veth_lpar_connection *cnx, | |
59f17aeb | 1402 | struct veth_lpevent *event) |
1da177e4 | 1403 | { |
59f17aeb | 1404 | struct veth_frames_data *senddata = &event->u.frames_data; |
1da177e4 LT |
1405 | int startchunk = 0; |
1406 | int nchunks; | |
1407 | unsigned long flags; | |
1408 | HvLpDma_Rc rc; | |
1409 | ||
1410 | do { | |
1411 | u16 length = 0; | |
1412 | struct sk_buff *skb; | |
1413 | struct dma_chunk local_list[VETH_MAX_PAGES_PER_FRAME]; | |
1414 | struct dma_chunk remote_list[VETH_MAX_FRAMES_PER_MSG]; | |
1415 | u64 dest; | |
1416 | HvLpVirtualLanIndex vlan; | |
1417 | struct net_device *dev; | |
1418 | struct veth_port *port; | |
1419 | ||
1420 | /* FIXME: do we need this? */ | |
1421 | memset(local_list, 0, sizeof(local_list)); | |
ba9bb431 | 1422 | memset(remote_list, 0, sizeof(remote_list)); |
1da177e4 LT |
1423 | |
1424 | /* a 0 address marks the end of the valid entries */ | |
1425 | if (senddata->addr[startchunk] == 0) | |
1426 | break; | |
1427 | ||
1428 | /* make sure that we have at least 1 EOF entry in the | |
1429 | * remaining entries */ | |
1430 | if (! (senddata->eofmask >> (startchunk + VETH_EOF_SHIFT))) { | |
61a3c696 ME |
1431 | veth_error("Missing EOF fragment in event " |
1432 | "eofmask = 0x%x startchunk = %d\n", | |
1433 | (unsigned)senddata->eofmask, | |
1434 | startchunk); | |
1da177e4 LT |
1435 | break; |
1436 | } | |
1437 | ||
1438 | /* build list of chunks in this frame */ | |
1439 | nchunks = 0; | |
1440 | do { | |
1441 | remote_list[nchunks].addr = | |
1442 | (u64) senddata->addr[startchunk+nchunks] << 32; | |
1443 | remote_list[nchunks].size = | |
1444 | senddata->len[startchunk+nchunks]; | |
1445 | length += remote_list[nchunks].size; | |
1446 | } while (! (senddata->eofmask & | |
1447 | (1 << (VETH_EOF_SHIFT + startchunk + nchunks++)))); | |
1448 | ||
1449 | /* length == total length of all chunks */ | |
1450 | /* nchunks == # of chunks in this frame */ | |
1451 | ||
1452 | if ((length - ETH_HLEN) > VETH_MAX_MTU) { | |
61a3c696 ME |
1453 | veth_error("Received oversize frame from LPAR %d " |
1454 | "(length = %d)\n", | |
1455 | cnx->remote_lp, length); | |
1da177e4 LT |
1456 | continue; |
1457 | } | |
1458 | ||
1459 | skb = alloc_skb(length, GFP_ATOMIC); | |
1460 | if (!skb) | |
1461 | continue; | |
1462 | ||
1463 | veth_build_dma_list(local_list, skb->data, length); | |
1464 | ||
1465 | rc = HvCallEvent_dmaBufList(HvLpEvent_Type_VirtualLan, | |
1466 | event->base_event.xSourceLp, | |
1467 | HvLpDma_Direction_RemoteToLocal, | |
1468 | cnx->src_inst, | |
1469 | cnx->dst_inst, | |
1470 | HvLpDma_AddressType_RealAddress, | |
1471 | HvLpDma_AddressType_TceIndex, | |
426c1a11 SR |
1472 | iseries_hv_addr(&local_list), |
1473 | iseries_hv_addr(&remote_list), | |
1da177e4 LT |
1474 | length); |
1475 | if (rc != HvLpDma_Rc_Good) { | |
1476 | dev_kfree_skb_irq(skb); | |
1477 | continue; | |
1478 | } | |
1479 | ||
1480 | vlan = skb->data[9]; | |
1481 | dev = veth_dev[vlan]; | |
41664c03 ME |
1482 | if (! dev) { |
1483 | /* | |
1484 | * Some earlier versions of the driver sent | |
1485 | * broadcasts down all connections, even to lpars | |
1486 | * that weren't on the relevant vlan. So ignore | |
1487 | * packets belonging to a vlan we're not on. | |
1488 | * We can also be here if we receive packets while | |
1489 | * the driver is going down, because then dev is NULL. | |
1490 | */ | |
1491 | dev_kfree_skb_irq(skb); | |
1da177e4 | 1492 | continue; |
41664c03 | 1493 | } |
1da177e4 | 1494 | |
454d7c9b | 1495 | port = netdev_priv(dev); |
1da177e4 LT |
1496 | dest = *((u64 *) skb->data) & 0xFFFFFFFFFFFF0000; |
1497 | ||
1498 | if ((vlan > HVMAXARCHITECTEDVIRTUALLANS) || !port) { | |
1499 | dev_kfree_skb_irq(skb); | |
1500 | continue; | |
1501 | } | |
1502 | if (! veth_frame_wanted(port, dest)) { | |
1503 | dev_kfree_skb_irq(skb); | |
1504 | continue; | |
1505 | } | |
1506 | ||
1507 | skb_put(skb, length); | |
1da177e4 | 1508 | skb->protocol = eth_type_trans(skb, dev); |
bc8acf2c | 1509 | skb_checksum_none_assert(skb); |
1da177e4 | 1510 | netif_rx(skb); /* send it up */ |
09f75cd7 JG |
1511 | dev->stats.rx_packets++; |
1512 | dev->stats.rx_bytes += length; | |
1da177e4 LT |
1513 | } while (startchunk += nchunks, startchunk < VETH_MAX_FRAMES_PER_MSG); |
1514 | ||
1515 | /* Ack it */ | |
1516 | spin_lock_irqsave(&cnx->lock, flags); | |
1517 | BUG_ON(cnx->num_pending_acks > VETH_MAX_ACKS_PER_MSG); | |
1518 | ||
1519 | cnx->pending_acks[cnx->num_pending_acks++] = | |
1520 | event->base_event.xCorrelationToken; | |
1521 | ||
8e95a202 JP |
1522 | if ( (cnx->num_pending_acks >= cnx->remote_caps.ack_threshold) || |
1523 | (cnx->num_pending_acks >= VETH_MAX_ACKS_PER_MSG) ) | |
1da177e4 LT |
1524 | veth_flush_acks(cnx); |
1525 | ||
1526 | spin_unlock_irqrestore(&cnx->lock, flags); | |
1527 | } | |
1528 | ||
1529 | static void veth_timed_ack(unsigned long ptr) | |
1530 | { | |
1531 | struct veth_lpar_connection *cnx = (struct veth_lpar_connection *) ptr; | |
1532 | unsigned long flags; | |
1533 | ||
1534 | /* Ack all the events */ | |
1535 | spin_lock_irqsave(&cnx->lock, flags); | |
1536 | if (cnx->num_pending_acks > 0) | |
1537 | veth_flush_acks(cnx); | |
1538 | ||
1539 | /* Reschedule the timer */ | |
1540 | cnx->ack_timer.expires = jiffies + cnx->ack_timeout; | |
1541 | add_timer(&cnx->ack_timer); | |
1542 | spin_unlock_irqrestore(&cnx->lock, flags); | |
1543 | } | |
1544 | ||
1545 | static int veth_remove(struct vio_dev *vdev) | |
1546 | { | |
f0c129ca | 1547 | struct veth_lpar_connection *cnx; |
1da177e4 | 1548 | struct net_device *dev; |
f0c129ca ME |
1549 | struct veth_port *port; |
1550 | int i; | |
1da177e4 | 1551 | |
f0c129ca ME |
1552 | dev = veth_dev[vdev->unit_address]; |
1553 | ||
1554 | if (! dev) | |
1555 | return 0; | |
1556 | ||
1557 | port = netdev_priv(dev); | |
1558 | ||
1559 | for (i = 0; i < HVMAXARCHITECTEDLPS; i++) { | |
1560 | cnx = veth_cnx[i]; | |
1561 | ||
1562 | if (cnx && (port->lpar_map & (1 << i))) { | |
1563 | /* Drop our reference to connections on our VLAN */ | |
1564 | kobject_put(&cnx->kobject); | |
1565 | } | |
1da177e4 | 1566 | } |
f0c129ca ME |
1567 | |
1568 | veth_dev[vdev->unit_address] = NULL; | |
07a5c172 ME |
1569 | kobject_del(&port->kobject); |
1570 | kobject_put(&port->kobject); | |
f0c129ca ME |
1571 | unregister_netdev(dev); |
1572 | free_netdev(dev); | |
1573 | ||
1da177e4 LT |
1574 | return 0; |
1575 | } | |
1576 | ||
1577 | static int veth_probe(struct vio_dev *vdev, const struct vio_device_id *id) | |
1578 | { | |
1579 | int i = vdev->unit_address; | |
1580 | struct net_device *dev; | |
f0c129ca | 1581 | struct veth_port *port; |
1da177e4 | 1582 | |
07fb3f45 | 1583 | dev = veth_probe_one(i, vdev); |
1da177e4 LT |
1584 | if (dev == NULL) { |
1585 | veth_remove(vdev); | |
1586 | return 1; | |
1587 | } | |
1588 | veth_dev[i] = dev; | |
1589 | ||
ece49153 | 1590 | port = netdev_priv(dev); |
f0c129ca ME |
1591 | |
1592 | /* Start the state machine on each connection on this vlan. If we're | |
1593 | * the first dev to do so this will commence link negotiation */ | |
1594 | for (i = 0; i < HVMAXARCHITECTEDLPS; i++) { | |
1595 | struct veth_lpar_connection *cnx; | |
1596 | ||
1597 | if (! (port->lpar_map & (1 << i))) | |
1598 | continue; | |
1599 | ||
1600 | cnx = veth_cnx[i]; | |
1601 | if (!cnx) | |
1602 | continue; | |
1603 | ||
1604 | kobject_get(&cnx->kobject); | |
1605 | veth_kick_statemachine(cnx); | |
1606 | } | |
1da177e4 LT |
1607 | |
1608 | return 0; | |
1609 | } | |
1610 | ||
1611 | /** | |
1612 | * veth_device_table: Used by vio.c to match devices that we | |
1613 | * support. | |
1614 | */ | |
1615 | static struct vio_device_id veth_device_table[] __devinitdata = { | |
07fb3f45 | 1616 | { "network", "IBM,iSeries-l-lan" }, |
fb120da6 | 1617 | { "", "" } |
1da177e4 LT |
1618 | }; |
1619 | MODULE_DEVICE_TABLE(vio, veth_device_table); | |
1620 | ||
1621 | static struct vio_driver veth_driver = { | |
1da177e4 LT |
1622 | .id_table = veth_device_table, |
1623 | .probe = veth_probe, | |
6fdf5392 SR |
1624 | .remove = veth_remove, |
1625 | .driver = { | |
1626 | .name = DRV_NAME, | |
915124d8 | 1627 | .owner = THIS_MODULE, |
6fdf5392 | 1628 | } |
1da177e4 LT |
1629 | }; |
1630 | ||
1631 | /* | |
1632 | * Module initialization/cleanup | |
1633 | */ | |
1634 | ||
687d18ab | 1635 | static void __exit veth_module_cleanup(void) |
1da177e4 LT |
1636 | { |
1637 | int i; | |
f0c129ca | 1638 | struct veth_lpar_connection *cnx; |
1da177e4 | 1639 | |
f0c129ca | 1640 | /* Disconnect our "irq" to stop events coming from the Hypervisor. */ |
1da177e4 LT |
1641 | HvLpEvent_unregisterHandler(HvLpEvent_Type_VirtualLan); |
1642 | ||
f0c129ca ME |
1643 | for (i = 0; i < HVMAXARCHITECTEDLPS; ++i) { |
1644 | cnx = veth_cnx[i]; | |
1645 | ||
1646 | if (!cnx) | |
1647 | continue; | |
b2e0852e | 1648 | |
9beb4896 TH |
1649 | /* Cancel work queued from Hypervisor callbacks */ |
1650 | cancel_delayed_work_sync(&cnx->statemachine_wq); | |
76812d81 ME |
1651 | /* Remove the connection from sysfs */ |
1652 | kobject_del(&cnx->kobject); | |
f0c129ca ME |
1653 | /* Drop the driver's reference to the connection */ |
1654 | kobject_put(&cnx->kobject); | |
1655 | } | |
1da177e4 | 1656 | |
f0c129ca ME |
1657 | /* Unregister the driver, which will close all the netdevs and stop |
1658 | * the connections when they're no longer referenced. */ | |
1659 | vio_unregister_driver(&veth_driver); | |
1da177e4 LT |
1660 | } |
1661 | module_exit(veth_module_cleanup); | |
1662 | ||
687d18ab | 1663 | static int __init veth_module_init(void) |
1da177e4 LT |
1664 | { |
1665 | int i; | |
1666 | int rc; | |
1667 | ||
687d18ab SR |
1668 | if (!firmware_has_feature(FW_FEATURE_ISERIES)) |
1669 | return -ENODEV; | |
1670 | ||
1da177e4 LT |
1671 | this_lp = HvLpConfig_getLpIndex_outline(); |
1672 | ||
1673 | for (i = 0; i < HVMAXARCHITECTEDLPS; ++i) { | |
1674 | rc = veth_init_connection(i); | |
ec60beeb ME |
1675 | if (rc != 0) |
1676 | goto error; | |
1da177e4 LT |
1677 | } |
1678 | ||
1679 | HvLpEvent_registerHandler(HvLpEvent_Type_VirtualLan, | |
1680 | &veth_handle_event); | |
1681 | ||
ec60beeb ME |
1682 | rc = vio_register_driver(&veth_driver); |
1683 | if (rc != 0) | |
1684 | goto error; | |
1685 | ||
76812d81 ME |
1686 | for (i = 0; i < HVMAXARCHITECTEDLPS; ++i) { |
1687 | struct kobject *kobj; | |
1688 | ||
1689 | if (!veth_cnx[i]) | |
1690 | continue; | |
1691 | ||
1692 | kobj = &veth_cnx[i]->kobject; | |
76812d81 | 1693 | /* If the add failes, complain but otherwise continue */ |
cbe9c595 | 1694 | if (0 != driver_add_kobj(&veth_driver.driver, kobj, |
78be3d2f | 1695 | "cnx%.2d", veth_cnx[i]->remote_lp)) |
76812d81 ME |
1696 | veth_error("cnx %d: Failed adding to sysfs.\n", i); |
1697 | } | |
1698 | ||
ec60beeb ME |
1699 | return 0; |
1700 | ||
1701 | error: | |
1702 | for (i = 0; i < HVMAXARCHITECTEDLPS; ++i) { | |
f0c129ca | 1703 | veth_destroy_connection(veth_cnx[i]); |
ec60beeb ME |
1704 | } |
1705 | ||
1706 | return rc; | |
1da177e4 LT |
1707 | } |
1708 | module_init(veth_module_init); |