1 |
1275 |
phoenix |
/*
|
2 |
|
|
* originally based on the dummy device.
|
3 |
|
|
*
|
4 |
|
|
* Copyright 1999, Thomas Davis, tadavis@lbl.gov.
|
5 |
|
|
* Licensed under the GPL. Based on dummy.c, and eql.c devices.
|
6 |
|
|
*
|
7 |
|
|
* bonding.c: an Ethernet Bonding driver
|
8 |
|
|
*
|
9 |
|
|
* This is useful to talk to a Cisco EtherChannel compatible equipment:
|
10 |
|
|
* Cisco 5500
|
11 |
|
|
* Sun Trunking (Solaris)
|
12 |
|
|
* Alteon AceDirector Trunks
|
13 |
|
|
* Linux Bonding
|
14 |
|
|
* and probably many L2 switches ...
|
15 |
|
|
*
|
16 |
|
|
* How it works:
|
17 |
|
|
* ifconfig bond0 ipaddress netmask up
|
18 |
|
|
* will setup a network device, with an ip address. No mac address
|
19 |
|
|
* will be assigned at this time. The hw mac address will come from
|
20 |
|
|
* the first slave bonded to the channel. All slaves will then use
|
21 |
|
|
* this hw mac address.
|
22 |
|
|
*
|
23 |
|
|
* ifconfig bond0 down
|
24 |
|
|
* will release all slaves, marking them as down.
|
25 |
|
|
*
|
26 |
|
|
* ifenslave bond0 eth0
|
27 |
|
|
* will attach eth0 to bond0 as a slave. eth0 hw mac address will either
|
28 |
|
|
* a: be used as initial mac address
|
29 |
|
|
* b: if a hw mac address already is there, eth0's hw mac address
|
30 |
|
|
* will then be set from bond0.
|
31 |
|
|
*
|
32 |
|
|
* v0.1 - first working version.
|
33 |
|
|
* v0.2 - changed stats to be calculated by summing slaves stats.
|
34 |
|
|
*
|
35 |
|
|
* Changes:
|
36 |
|
|
* Arnaldo Carvalho de Melo <acme@conectiva.com.br>
|
37 |
|
|
* - fix leaks on failure at bond_init
|
38 |
|
|
*
|
39 |
|
|
* 2000/09/30 - Willy Tarreau <willy at meta-x.org>
|
40 |
|
|
* - added trivial code to release a slave device.
|
41 |
|
|
* - fixed security bug (CAP_NET_ADMIN not checked)
|
42 |
|
|
* - implemented MII link monitoring to disable dead links :
|
43 |
|
|
* All MII capable slaves are checked every <miimon> milliseconds
|
44 |
|
|
* (100 ms seems good). This value can be changed by passing it to
|
45 |
|
|
* insmod. A value of zero disables the monitoring (default).
|
46 |
|
|
* - fixed an infinite loop in bond_xmit_roundrobin() when there's no
|
47 |
|
|
* good slave.
|
48 |
|
|
* - made the code hopefully SMP safe
|
49 |
|
|
*
|
50 |
|
|
* 2000/10/03 - Willy Tarreau <willy at meta-x.org>
|
51 |
|
|
* - optimized slave lists based on relevant suggestions from Thomas Davis
|
52 |
|
|
* - implemented active-backup method to obtain HA with two switches:
|
53 |
|
|
* stay as long as possible on the same active interface, while we
|
54 |
|
|
* also monitor the backup one (MII link status) because we want to know
|
55 |
|
|
* if we are able to switch at any time. ( pass "mode=1" to insmod )
|
56 |
|
|
* - lots of stress testings because we need it to be more robust than the
|
57 |
|
|
* wires ! :->
|
58 |
|
|
*
|
59 |
|
|
* 2000/10/09 - Willy Tarreau <willy at meta-x.org>
|
60 |
|
|
* - added up and down delays after link state change.
|
61 |
|
|
* - optimized the slaves chaining so that when we run forward, we never
|
62 |
|
|
* repass through the bond itself, but we can find it by searching
|
63 |
|
|
* backwards. Renders the deletion more difficult, but accelerates the
|
64 |
|
|
* scan.
|
65 |
|
|
* - smarter enslaving and releasing.
|
66 |
|
|
* - finer and more robust SMP locking
|
67 |
|
|
*
|
68 |
|
|
* 2000/10/17 - Willy Tarreau <willy at meta-x.org>
|
69 |
|
|
* - fixed two potential SMP race conditions
|
70 |
|
|
*
|
71 |
|
|
* 2000/10/18 - Willy Tarreau <willy at meta-x.org>
|
72 |
|
|
* - small fixes to the monitoring FSM in case of zero delays
|
73 |
|
|
* 2000/11/01 - Willy Tarreau <willy at meta-x.org>
|
74 |
|
|
* - fixed first slave not automatically used in trunk mode.
|
75 |
|
|
* 2000/11/10 : spelling of "EtherChannel" corrected.
|
76 |
|
|
* 2000/11/13 : fixed a race condition in case of concurrent accesses to ioctl().
|
77 |
|
|
* 2000/12/16 : fixed improper usage of rtnl_exlock_nowait().
|
78 |
|
|
*
|
79 |
|
|
* 2001/1/3 - Chad N. Tindel <ctindel at ieee dot org>
|
80 |
|
|
* - The bonding driver now simulates MII status monitoring, just like
|
81 |
|
|
* a normal network device. It will show that the link is down iff
|
82 |
|
|
* every slave in the bond shows that their links are down. If at least
|
83 |
|
|
* one slave is up, the bond's MII status will appear as up.
|
84 |
|
|
*
|
85 |
|
|
* 2001/2/7 - Chad N. Tindel <ctindel at ieee dot org>
|
86 |
|
|
* - Applications can now query the bond from user space to get
|
87 |
|
|
* information which may be useful. They do this by calling
|
88 |
|
|
* the BOND_INFO_QUERY ioctl. Once the app knows how many slaves
|
89 |
|
|
* are in the bond, it can call the BOND_SLAVE_INFO_QUERY ioctl to
|
90 |
|
|
* get slave specific information (# link failures, etc). See
|
91 |
|
|
* <linux/if_bonding.h> for more details. The structs of interest
|
92 |
|
|
* are ifbond and ifslave.
|
93 |
|
|
*
|
94 |
|
|
* 2001/4/5 - Chad N. Tindel <ctindel at ieee dot org>
|
95 |
|
|
* - Ported to 2.4 Kernel
|
96 |
|
|
*
|
97 |
|
|
* 2001/5/2 - Jeffrey E. Mast <jeff at mastfamily dot com>
|
98 |
|
|
* - When a device is detached from a bond, the slave device is no longer
|
99 |
|
|
* left thinking that is has a master.
|
100 |
|
|
*
|
101 |
|
|
* 2001/5/16 - Jeffrey E. Mast <jeff at mastfamily dot com>
|
102 |
|
|
* - memset did not appropriately initialized the bond rw_locks. Used
|
103 |
|
|
* rwlock_init to initialize to unlocked state to prevent deadlock when
|
104 |
|
|
* first attempting a lock
|
105 |
|
|
* - Called SET_MODULE_OWNER for bond device
|
106 |
|
|
*
|
107 |
|
|
* 2001/5/17 - Tim Anderson <tsa at mvista.com>
|
108 |
|
|
* - 2 paths for releasing for slave release; 1 through ioctl
|
109 |
|
|
* and 2) through close. Both paths need to release the same way.
|
110 |
|
|
* - the free slave in bond release is changing slave status before
|
111 |
|
|
* the free. The netdev_set_master() is intended to change slave state
|
112 |
|
|
* so it should not be done as part of the release process.
|
113 |
|
|
* - Simple rule for slave state at release: only the active in A/B and
|
114 |
|
|
* only one in the trunked case.
|
115 |
|
|
*
|
116 |
|
|
* 2001/6/01 - Tim Anderson <tsa at mvista.com>
|
117 |
|
|
* - Now call dev_close when releasing a slave so it doesn't screw up
|
118 |
|
|
* out routing table.
|
119 |
|
|
*
|
120 |
|
|
* 2001/6/01 - Chad N. Tindel <ctindel at ieee dot org>
|
121 |
|
|
* - Added /proc support for getting bond and slave information.
|
122 |
|
|
* Information is in /proc/net/<bond device>/info.
|
123 |
|
|
* - Changed the locking when calling bond_close to prevent deadlock.
|
124 |
|
|
*
|
125 |
|
|
* 2001/8/05 - Janice Girouard <girouard at us.ibm.com>
|
126 |
|
|
* - correct problem where refcnt of slave is not incremented in bond_ioctl
|
127 |
|
|
* so the system hangs when halting.
|
128 |
|
|
* - correct locking problem when unable to malloc in bond_enslave.
|
129 |
|
|
* - adding bond_xmit_xor logic.
|
130 |
|
|
* - adding multiple bond device support.
|
131 |
|
|
*
|
132 |
|
|
* 2001/8/13 - Erik Habbinga <erik_habbinga at hp dot com>
|
133 |
|
|
* - correct locking problem with rtnl_exlock_nowait
|
134 |
|
|
*
|
135 |
|
|
* 2001/8/23 - Janice Girouard <girouard at us.ibm.com>
|
136 |
|
|
* - bzero initial dev_bonds, to correct oops
|
137 |
|
|
* - convert SIOCDEVPRIVATE to new MII ioctl calls
|
138 |
|
|
*
|
139 |
|
|
* 2001/9/13 - Takao Indoh <indou dot takao at jp dot fujitsu dot com>
|
140 |
|
|
* - Add the BOND_CHANGE_ACTIVE ioctl implementation
|
141 |
|
|
*
|
142 |
|
|
* 2001/9/14 - Mark Huth <mhuth at mvista dot com>
|
143 |
|
|
* - Change MII_LINK_READY to not check for end of auto-negotiation,
|
144 |
|
|
* but only for an up link.
|
145 |
|
|
*
|
146 |
|
|
* 2001/9/20 - Chad N. Tindel <ctindel at ieee dot org>
|
147 |
|
|
* - Add the device field to bonding_t. Previously the net_device
|
148 |
|
|
* corresponding to a bond wasn't available from the bonding_t
|
149 |
|
|
* structure.
|
150 |
|
|
*
|
151 |
|
|
* 2001/9/25 - Janice Girouard <girouard at us.ibm.com>
|
152 |
|
|
* - add arp_monitor for active backup mode
|
153 |
|
|
*
|
154 |
|
|
* 2001/10/23 - Takao Indoh <indou dot takao at jp dot fujitsu dot com>
|
155 |
|
|
* - Various memory leak fixes
|
156 |
|
|
*
|
157 |
|
|
* 2001/11/5 - Mark Huth <mark dot huth at mvista dot com>
|
158 |
|
|
* - Don't take rtnl lock in bond_mii_monitor as it deadlocks under
|
159 |
|
|
* certain hotswap conditions.
|
160 |
|
|
* Note: this same change may be required in bond_arp_monitor ???
|
161 |
|
|
* - Remove possibility of calling bond_sethwaddr with NULL slave_dev ptr
|
162 |
|
|
* - Handle hot swap ethernet interface deregistration events to remove
|
163 |
|
|
* kernel oops following hot swap of enslaved interface
|
164 |
|
|
*
|
165 |
|
|
* 2002/1/2 - Chad N. Tindel <ctindel at ieee dot org>
|
166 |
|
|
* - Restore original slave flags at release time.
|
167 |
|
|
*
|
168 |
|
|
* 2002/02/18 - Erik Habbinga <erik_habbinga at hp dot com>
|
169 |
|
|
* - bond_release(): calling kfree on our_slave after call to
|
170 |
|
|
* bond_restore_slave_flags, not before
|
171 |
|
|
* - bond_enslave(): saving slave flags into original_flags before
|
172 |
|
|
* call to netdev_set_master, so the IFF_SLAVE flag doesn't end
|
173 |
|
|
* up in original_flags
|
174 |
|
|
*
|
175 |
|
|
* 2002/04/05 - Mark Smith <mark.smith at comdev dot cc> and
|
176 |
|
|
* Steve Mead <steve.mead at comdev dot cc>
|
177 |
|
|
* - Port Gleb Natapov's multicast support patchs from 2.4.12
|
178 |
|
|
* to 2.4.18 adding support for multicast.
|
179 |
|
|
*
|
180 |
|
|
* 2002/06/10 - Tony Cureington <tony.cureington * hp_com>
|
181 |
|
|
* - corrected uninitialized pointer (ifr.ifr_data) in bond_check_dev_link;
|
182 |
|
|
* actually changed function to use MIIPHY, then MIIREG, and finally
|
183 |
|
|
* ETHTOOL to determine the link status
|
184 |
|
|
* - fixed bad ifr_data pointer assignments in bond_ioctl
|
185 |
|
|
* - corrected mode 1 being reported as active-backup in bond_get_info;
|
186 |
|
|
* also added text to distinguish type of load balancing (rr or xor)
|
187 |
|
|
* - change arp_ip_target module param from "1-12s" (array of 12 ptrs)
|
188 |
|
|
* to "s" (a single ptr)
|
189 |
|
|
*
|
190 |
|
|
* 2002/08/30 - Jay Vosburgh <fubar at us dot ibm dot com>
|
191 |
|
|
* - Removed acquisition of xmit_lock in set_multicast_list; caused
|
192 |
|
|
* deadlock on SMP (lock is held by caller).
|
193 |
|
|
* - Revamped SIOCGMIIPHY, SIOCGMIIREG portion of bond_check_dev_link().
|
194 |
|
|
*
|
195 |
|
|
* 2002/09/18 - Jay Vosburgh <fubar at us dot ibm dot com>
|
196 |
|
|
* - Fixed up bond_check_dev_link() (and callers): removed some magic
|
197 |
|
|
* numbers, banished local MII_ defines, wrapped ioctl calls to
|
198 |
|
|
* prevent EFAULT errors
|
199 |
|
|
*
|
200 |
|
|
* 2002/9/30 - Jay Vosburgh <fubar at us dot ibm dot com>
|
201 |
|
|
* - make sure the ip target matches the arp_target before saving the
|
202 |
|
|
* hw address.
|
203 |
|
|
*
|
204 |
|
|
* 2002/9/30 - Dan Eisner <eisner at 2robots dot com>
|
205 |
|
|
* - make sure my_ip is set before taking down the link, since
|
206 |
|
|
* not all switches respond if the source ip is not set.
|
207 |
|
|
*
|
208 |
|
|
* 2002/10/8 - Janice Girouard <girouard at us dot ibm dot com>
|
209 |
|
|
* - read in the local ip address when enslaving a device
|
210 |
|
|
* - add primary support
|
211 |
|
|
* - make sure 2*arp_interval has passed when a new device
|
212 |
|
|
* is brought on-line before taking it down.
|
213 |
|
|
*
|
214 |
|
|
* 2002/09/11 - Philippe De Muyter <phdm at macqel dot be>
|
215 |
|
|
* - Added bond_xmit_broadcast logic.
|
216 |
|
|
* - Added bond_mode() support function.
|
217 |
|
|
*
|
218 |
|
|
* 2002/10/26 - Laurent Deniel <laurent.deniel at free.fr>
|
219 |
|
|
* - allow to register multicast addresses only on active slave
|
220 |
|
|
* (useful in active-backup mode)
|
221 |
|
|
* - add multicast module parameter
|
222 |
|
|
* - fix deletion of multicast groups after unloading module
|
223 |
|
|
*
|
224 |
|
|
* 2002/11/06 - Kameshwara Rayaprolu <kameshwara.rao * wipro_com>
|
225 |
|
|
* - Changes to prevent panic from closing the device twice; if we close
|
226 |
|
|
* the device in bond_release, we must set the original_flags to down
|
227 |
|
|
* so it won't be closed again by the network layer.
|
228 |
|
|
*
|
229 |
|
|
* 2002/11/07 - Tony Cureington <tony.cureington * hp_com>
|
230 |
|
|
* - Fix arp_target_hw_addr memory leak
|
231 |
|
|
* - Created activebackup_arp_monitor function to handle arp monitoring
|
232 |
|
|
* in active backup mode - the bond_arp_monitor had several problems...
|
233 |
|
|
* such as allowing slaves to tx arps sequentially without any delay
|
234 |
|
|
* for a response
|
235 |
|
|
* - Renamed bond_arp_monitor to loadbalance_arp_monitor and re-wrote
|
236 |
|
|
* this function to just handle arp monitoring in load-balancing mode;
|
237 |
|
|
* it is a lot more compact now
|
238 |
|
|
* - Changes to ensure one and only one slave transmits in active-backup
|
239 |
|
|
* mode
|
240 |
|
|
* - Robustesize parameters; warn users about bad combinations of
|
241 |
|
|
* parameters; also if miimon is specified and a network driver does
|
242 |
|
|
* not support MII or ETHTOOL, inform the user of this
|
243 |
|
|
* - Changes to support link_failure_count when in arp monitoring mode
|
244 |
|
|
* - Fix up/down delay reported in /proc
|
245 |
|
|
* - Added version; log version; make version available from "modinfo -d"
|
246 |
|
|
* - Fixed problem in bond_check_dev_link - if the first IOCTL (SIOCGMIIPH)
|
247 |
|
|
* failed, the ETHTOOL ioctl never got a chance
|
248 |
|
|
*
|
249 |
|
|
* 2002/11/16 - Laurent Deniel <laurent.deniel at free.fr>
|
250 |
|
|
* - fix multicast handling in activebackup_arp_monitor
|
251 |
|
|
* - remove one unnecessary and confusing curr_active_slave == slave test
|
252 |
|
|
* in activebackup_arp_monitor
|
253 |
|
|
*
|
254 |
|
|
* 2002/11/17 - Laurent Deniel <laurent.deniel at free.fr>
|
255 |
|
|
* - fix bond_slave_info_query when slave_id = num_slaves
|
256 |
|
|
*
|
257 |
|
|
* 2002/11/19 - Janice Girouard <girouard at us dot ibm dot com>
|
258 |
|
|
* - correct ifr_data reference. Update ifr_data reference
|
259 |
|
|
* to mii_ioctl_data struct values to avoid confusion.
|
260 |
|
|
*
|
261 |
|
|
* 2002/11/22 - Bert Barbe <bert.barbe at oracle dot com>
|
262 |
|
|
* - Add support for multiple arp_ip_target
|
263 |
|
|
*
|
264 |
|
|
* 2002/12/13 - Jay Vosburgh <fubar at us dot ibm dot com>
|
265 |
|
|
* - Changed to allow text strings for mode and multicast, e.g.,
|
266 |
|
|
* insmod bonding mode=active-backup. The numbers still work.
|
267 |
|
|
* One change: an invalid choice will cause module load failure,
|
268 |
|
|
* rather than the previous behavior of just picking one.
|
269 |
|
|
* - Minor cleanups; got rid of dup ctype stuff, atoi function
|
270 |
|
|
*
|
271 |
|
|
* 2003/02/07 - Jay Vosburgh <fubar at us dot ibm dot com>
|
272 |
|
|
* - Added use_carrier module parameter that causes miimon to
|
273 |
|
|
* use netif_carrier_ok() test instead of MII/ETHTOOL ioctls.
|
274 |
|
|
* - Minor cleanups; consolidated ioctl calls to one function.
|
275 |
|
|
*
|
276 |
|
|
* 2003/02/07 - Tony Cureington <tony.cureington * hp_com>
|
277 |
|
|
* - Fix bond_mii_monitor() logic error that could result in
|
278 |
|
|
* bonding round-robin mode ignoring links after failover/recovery
|
279 |
|
|
*
|
280 |
|
|
* 2003/03/17 - Jay Vosburgh <fubar at us dot ibm dot com>
|
281 |
|
|
* - kmalloc fix (GFP_KERNEL to GFP_ATOMIC) reported by
|
282 |
|
|
* Shmulik dot Hen at intel.com.
|
283 |
|
|
* - Based on discussion on mailing list, changed use of
|
284 |
|
|
* update_slave_cnt(), created wrapper functions for adding/removing
|
285 |
|
|
* slaves, changed bond_xmit_xor() to check slave_cnt instead of
|
286 |
|
|
* checking slave and slave->dev (which only worked by accident).
|
287 |
|
|
* - Misc code cleanup: get arp_send() prototype from header file,
|
288 |
|
|
* add max_bonds to bonding.txt.
|
289 |
|
|
*
|
290 |
|
|
* 2003/03/18 - Tsippy Mendelson <tsippy.mendelson at intel dot com> and
|
291 |
|
|
* Shmulik Hen <shmulik.hen at intel dot com>
|
292 |
|
|
* - Make sure only bond_attach_slave() and bond_detach_slave() can
|
293 |
|
|
* manipulate the slave list, including slave_cnt, even when in
|
294 |
|
|
* bond_release_all().
|
295 |
|
|
* - Fixed hang in bond_release() with traffic running:
|
296 |
|
|
* netdev_set_master() must not be called from within the bond lock.
|
297 |
|
|
*
|
298 |
|
|
* 2003/03/18 - Tsippy Mendelson <tsippy.mendelson at intel dot com> and
|
299 |
|
|
* Shmulik Hen <shmulik.hen at intel dot com>
|
300 |
|
|
* - Fixed hang in bond_enslave() with traffic running:
|
301 |
|
|
* netdev_set_master() must not be called from within the bond lock.
|
302 |
|
|
*
|
303 |
|
|
* 2003/03/18 - Amir Noam <amir.noam at intel dot com>
|
304 |
|
|
* - Added support for getting slave's speed and duplex via ethtool.
|
305 |
|
|
* Needed for 802.3ad and other future modes.
|
306 |
|
|
*
|
307 |
|
|
* 2003/03/18 - Tsippy Mendelson <tsippy.mendelson at intel dot com> and
|
308 |
|
|
* Shmulik Hen <shmulik.hen at intel dot com>
|
309 |
|
|
* - Enable support of modes that need to use the unique mac address of
|
310 |
|
|
* each slave.
|
311 |
|
|
* * bond_enslave(): Moved setting the slave's mac address, and
|
312 |
|
|
* openning it, from the application to the driver. This breaks
|
313 |
|
|
* backward comaptibility with old versions of ifenslave that open
|
314 |
|
|
* the slave before enalsving it !!!.
|
315 |
|
|
* * bond_release(): The driver also takes care of closing the slave
|
316 |
|
|
* and restoring its original mac address.
|
317 |
|
|
* - Removed the code that restores all base driver's flags.
|
318 |
|
|
* Flags are automatically restored once all undo stages are done
|
319 |
|
|
* properly.
|
320 |
|
|
* - Block possibility of enslaving before the master is up. This
|
321 |
|
|
* prevents putting the system in an unstable state.
|
322 |
|
|
*
|
323 |
|
|
* 2003/03/18 - Amir Noam <amir.noam at intel dot com>,
|
324 |
|
|
* Tsippy Mendelson <tsippy.mendelson at intel dot com> and
|
325 |
|
|
* Shmulik Hen <shmulik.hen at intel dot com>
|
326 |
|
|
* - Added support for IEEE 802.3ad Dynamic link aggregation mode.
|
327 |
|
|
*
|
328 |
|
|
* 2003/05/01 - Amir Noam <amir.noam at intel dot com>
|
329 |
|
|
* - Added ABI version control to restore compatibility between
|
330 |
|
|
* new/old ifenslave and new/old bonding.
|
331 |
|
|
*
|
332 |
|
|
* 2003/05/01 - Shmulik Hen <shmulik.hen at intel dot com>
|
333 |
|
|
* - Fixed bug in bond_release_all(): save old value of curr_active_slave
|
334 |
|
|
* before setting it to NULL.
|
335 |
|
|
* - Changed driver versioning scheme to include version number instead
|
336 |
|
|
* of release date (that is already in another field). There are 3
|
337 |
|
|
* fields X.Y.Z where:
|
338 |
|
|
* X - Major version - big behavior changes
|
339 |
|
|
* Y - Minor version - addition of features
|
340 |
|
|
* Z - Extra version - minor changes and bug fixes
|
341 |
|
|
* The current version is 1.0.0 as a base line.
|
342 |
|
|
*
|
343 |
|
|
* 2003/05/01 - Tsippy Mendelson <tsippy.mendelson at intel dot com> and
|
344 |
|
|
* Amir Noam <amir.noam at intel dot com>
|
345 |
|
|
* - Added support for lacp_rate module param.
|
346 |
|
|
* - Code beautification and style changes (mainly in comments).
|
347 |
|
|
* new version - 1.0.1
|
348 |
|
|
*
|
349 |
|
|
* 2003/05/01 - Shmulik Hen <shmulik.hen at intel dot com>
|
350 |
|
|
* - Based on discussion on mailing list, changed locking scheme
|
351 |
|
|
* to use lock/unlock or lock_bh/unlock_bh appropriately instead
|
352 |
|
|
* of lock_irqsave/unlock_irqrestore. The new scheme helps exposing
|
353 |
|
|
* hidden bugs and solves system hangs that occurred due to the fact
|
354 |
|
|
* that holding lock_irqsave doesn't prevent softirqs from running.
|
355 |
|
|
* This also increases total throughput since interrupts are not
|
356 |
|
|
* blocked on each transmitted packets or monitor timeout.
|
357 |
|
|
* new version - 2.0.0
|
358 |
|
|
*
|
359 |
|
|
* 2003/05/01 - Shmulik Hen <shmulik.hen at intel dot com>
|
360 |
|
|
* - Added support for Transmit load balancing mode.
|
361 |
|
|
* - Concentrate all assignments of curr_active_slave to a single point
|
362 |
|
|
* so specific modes can take actions when the primary adapter is
|
363 |
|
|
* changed.
|
364 |
|
|
* - Take the updelay parameter into consideration during bond_enslave
|
365 |
|
|
* since some adapters loose their link during setting the device.
|
366 |
|
|
* - Renamed bond_3ad_link_status_changed() to
|
367 |
|
|
* bond_3ad_handle_link_change() for compatibility with TLB.
|
368 |
|
|
* new version - 2.1.0
|
369 |
|
|
*
|
370 |
|
|
* 2003/05/01 - Tsippy Mendelson <tsippy.mendelson at intel dot com>
|
371 |
|
|
* - Added support for Adaptive load balancing mode which is
|
372 |
|
|
* equivalent to Transmit load balancing + Receive load balancing.
|
373 |
|
|
* new version - 2.2.0
|
374 |
|
|
*
|
375 |
|
|
* 2003/05/15 - Jay Vosburgh <fubar at us dot ibm dot com>
|
376 |
|
|
* - Applied fix to activebackup_arp_monitor posted to bonding-devel
|
377 |
|
|
* by Tony Cureington <tony.cureington * hp_com>. Fixes ARP
|
378 |
|
|
* monitor endless failover bug. Version to 2.2.10
|
379 |
|
|
*
|
380 |
|
|
* 2003/05/20 - Amir Noam <amir.noam at intel dot com>
|
381 |
|
|
* - Fixed bug in ABI version control - Don't commit to a specific
|
382 |
|
|
* ABI version if receiving unsupported ioctl commands.
|
383 |
|
|
*
|
384 |
|
|
* 2003/05/22 - Jay Vosburgh <fubar at us dot ibm dot com>
|
385 |
|
|
* - Fix ifenslave -c causing bond to loose existing routes;
|
386 |
|
|
* added bond_set_mac_address() that doesn't require the
|
387 |
|
|
* bond to be down.
|
388 |
|
|
* - In conjunction with fix for ifenslave -c, in
|
389 |
|
|
* bond_change_active(), changing to the already active slave
|
390 |
|
|
* is no longer an error (it successfully does nothing).
|
391 |
|
|
*
|
392 |
|
|
* 2003/06/30 - Amir Noam <amir.noam at intel dot com>
|
393 |
|
|
* - Fixed bond_change_active() for ALB/TLB modes.
|
394 |
|
|
* Version to 2.2.14.
|
395 |
|
|
*
|
396 |
|
|
* 2003/07/29 - Amir Noam <amir.noam at intel dot com>
|
397 |
|
|
* - Fixed ARP monitoring bug.
|
398 |
|
|
* Version to 2.2.15.
|
399 |
|
|
*
|
400 |
|
|
* 2003/07/31 - Willy Tarreau <willy at ods dot org>
|
401 |
|
|
* - Fixed kernel panic when using ARP monitoring without
|
402 |
|
|
* setting bond's IP address.
|
403 |
|
|
* Version to 2.2.16.
|
404 |
|
|
*
|
405 |
|
|
* 2003/08/06 - Amir Noam <amir.noam at intel dot com>
|
406 |
|
|
* - Back port from 2.6: use alloc_netdev(); fix /proc handling;
|
407 |
|
|
* made stats a part of bond struct so no need to allocate
|
408 |
|
|
* and free it separately; use standard list operations instead
|
409 |
|
|
* of pre-allocated array of bonds.
|
410 |
|
|
* Version to 2.3.0.
|
411 |
|
|
*
|
412 |
|
|
* 2003/08/07 - Jay Vosburgh <fubar at us dot ibm dot com>,
|
413 |
|
|
* Amir Noam <amir.noam at intel dot com> and
|
414 |
|
|
* Shmulik Hen <shmulik.hen at intel dot com>
|
415 |
|
|
* - Propagating master's settings: Distinguish between modes that
|
416 |
|
|
* use a primary slave from those that don't, and propagate settings
|
417 |
|
|
* accordingly; Consolidate change_active opeartions and add
|
418 |
|
|
* reselect_active and find_best opeartions; Decouple promiscuous
|
419 |
|
|
* handling from the multicast mode setting; Add support for changing
|
420 |
|
|
* HW address and MTU with proper unwind; Consolidate procfs code,
|
421 |
|
|
* add CHANGENAME handler; Enhance netdev notification handling.
|
422 |
|
|
* Version to 2.4.0.
|
423 |
|
|
*
|
424 |
|
|
* 2003/09/15 - Stephen Hemminger <shemminger at osdl dot org>,
|
425 |
|
|
* Amir Noam <amir.noam at intel dot com>
|
426 |
|
|
* - Convert /proc to seq_file interface.
|
427 |
|
|
* Change /proc/net/bondX/info to /proc/net/bonding/bondX.
|
428 |
|
|
* Set version to 2.4.1.
|
429 |
|
|
*
|
430 |
|
|
* 2003/11/20 - Amir Noam <amir.noam at intel dot com>
|
431 |
|
|
* - Fix /proc creation/destruction.
|
432 |
|
|
*
|
433 |
|
|
* 2003/12/01 - Shmulik Hen <shmulik.hen at intel dot com>
|
434 |
|
|
* - Massive cleanup - Set version to 2.5.0
|
435 |
|
|
* Code changes:
|
436 |
|
|
* o Consolidate format of prints and debug prints.
|
437 |
|
|
* o Remove bonding_t/slave_t typedefs and consolidate all casts.
|
438 |
|
|
* o Remove dead code and unnecessary checks.
|
439 |
|
|
* o Consolidate starting/stopping timers.
|
440 |
|
|
* o Consolidate handling of primary module param throughout the code.
|
441 |
|
|
* o Removed multicast module param support - all settings are done
|
442 |
|
|
* according to mode.
|
443 |
|
|
* o Slave list iteration - bond is no longer part of the list,
|
444 |
|
|
* added cyclic list iteration macros.
|
445 |
|
|
* o Consolidate error handling in all xmit functions.
|
446 |
|
|
* Style changes:
|
447 |
|
|
* o Consolidate function naming and declarations.
|
448 |
|
|
* o Consolidate function params and local variables names.
|
449 |
|
|
* o Consolidate return values.
|
450 |
|
|
* o Consolidate curly braces.
|
451 |
|
|
* o Consolidate conditionals format.
|
452 |
|
|
* o Change struct member names and types.
|
453 |
|
|
* o Chomp trailing spaces, remove empty lines, fix indentations.
|
454 |
|
|
* o Re-organize code according to context.
|
455 |
|
|
*
|
456 |
|
|
* 2003/12/30 - Amir Noam <amir.noam at intel dot com>
|
457 |
|
|
* - Fixed: Cannot remove and re-enslave the original active slave.
|
458 |
|
|
* - Fixed: Releasing the original active slave causes mac address
|
459 |
|
|
* duplication.
|
460 |
|
|
* - Add support for slaves that use ethtool_ops.
|
461 |
|
|
* Set version to 2.5.3.
|
462 |
|
|
*
|
463 |
|
|
* 2004/01/05 - Amir Noam <amir.noam at intel dot com>
|
464 |
|
|
* - Save bonding parameters per bond instead of using the global values.
|
465 |
|
|
* Set version to 2.5.4.
|
466 |
|
|
*
|
467 |
|
|
* 2004/01/14 - Shmulik Hen <shmulik.hen at intel dot com>
|
468 |
|
|
* - Enhance VLAN support:
|
469 |
|
|
* * Add support for VLAN hardware acceleration capable slaves.
|
470 |
|
|
* * Add capability to tag self generated packets in ALB/TLB modes.
|
471 |
|
|
* Set version to 2.6.0.
|
472 |
|
|
*/
|
473 |
|
|
|
474 |
|
|
//#define BONDING_DEBUG 1
|
475 |
|
|
|
476 |
|
|
#include <linux/config.h>
|
477 |
|
|
#include <linux/kernel.h>
|
478 |
|
|
#include <linux/module.h>
|
479 |
|
|
#include <linux/sched.h>
|
480 |
|
|
#include <linux/types.h>
|
481 |
|
|
#include <linux/fcntl.h>
|
482 |
|
|
#include <linux/interrupt.h>
|
483 |
|
|
#include <linux/ptrace.h>
|
484 |
|
|
#include <linux/ioport.h>
|
485 |
|
|
#include <linux/in.h>
|
486 |
|
|
#include <linux/ip.h>
|
487 |
|
|
#include <linux/slab.h>
|
488 |
|
|
#include <linux/string.h>
|
489 |
|
|
#include <linux/init.h>
|
490 |
|
|
#include <linux/timer.h>
|
491 |
|
|
#include <linux/socket.h>
|
492 |
|
|
#include <linux/ctype.h>
|
493 |
|
|
#include <linux/inet.h>
|
494 |
|
|
#include <asm/system.h>
|
495 |
|
|
#include <asm/bitops.h>
|
496 |
|
|
#include <asm/io.h>
|
497 |
|
|
#include <asm/dma.h>
|
498 |
|
|
#include <asm/uaccess.h>
|
499 |
|
|
#include <linux/errno.h>
|
500 |
|
|
#include <linux/netdevice.h>
|
501 |
|
|
#include <linux/inetdevice.h>
|
502 |
|
|
#include <linux/etherdevice.h>
|
503 |
|
|
#include <linux/skbuff.h>
|
504 |
|
|
#include <net/sock.h>
|
505 |
|
|
#include <linux/rtnetlink.h>
|
506 |
|
|
#include <linux/proc_fs.h>
|
507 |
|
|
#include <linux/seq_file.h>
|
508 |
|
|
#include <linux/smp.h>
|
509 |
|
|
#include <linux/if_ether.h>
|
510 |
|
|
#include <net/arp.h>
|
511 |
|
|
#include <linux/mii.h>
|
512 |
|
|
#include <linux/ethtool.h>
|
513 |
|
|
#include <linux/if_vlan.h>
|
514 |
|
|
#include <linux/if_bonding.h>
|
515 |
|
|
#include "bonding.h"
|
516 |
|
|
#include "bond_3ad.h"
|
517 |
|
|
#include "bond_alb.h"
|
518 |
|
|
|
519 |
|
|
/*---------------------------- Module parameters ----------------------------*/
|
520 |
|
|
|
521 |
|
|
/* monitor all links that often (in milliseconds). <=0 disables monitoring */
|
522 |
|
|
#define BOND_LINK_MON_INTERV 0
|
523 |
|
|
#define BOND_LINK_ARP_INTERV 0
|
524 |
|
|
|
525 |
|
|
static int max_bonds = BOND_DEFAULT_MAX_BONDS;
|
526 |
|
|
static int miimon = BOND_LINK_MON_INTERV;
|
527 |
|
|
static int updelay = 0;
|
528 |
|
|
static int downdelay = 0;
|
529 |
|
|
static int use_carrier = 1;
|
530 |
|
|
static char *mode = NULL;
|
531 |
|
|
static char *primary = NULL;
|
532 |
|
|
static char *lacp_rate = NULL;
|
533 |
|
|
static int arp_interval = BOND_LINK_ARP_INTERV;
|
534 |
|
|
static char *arp_ip_target[BOND_MAX_ARP_TARGETS] = { NULL, };
|
535 |
|
|
|
536 |
|
|
MODULE_PARM(max_bonds, "i");
|
537 |
|
|
MODULE_PARM_DESC(max_bonds, "Max number of bonded devices");
|
538 |
|
|
MODULE_PARM(miimon, "i");
|
539 |
|
|
MODULE_PARM_DESC(miimon, "Link check interval in milliseconds");
|
540 |
|
|
MODULE_PARM(updelay, "i");
|
541 |
|
|
MODULE_PARM_DESC(updelay, "Delay before considering link up, in milliseconds");
|
542 |
|
|
MODULE_PARM(downdelay, "i");
|
543 |
|
|
MODULE_PARM_DESC(downdelay, "Delay before considering link down, in milliseconds");
|
544 |
|
|
MODULE_PARM(use_carrier, "i");
|
545 |
|
|
MODULE_PARM_DESC(use_carrier, "Use netif_carrier_ok (vs MII ioctls) in miimon; 0 for off, 1 for on (default)");
|
546 |
|
|
MODULE_PARM(mode, "s");
|
547 |
|
|
MODULE_PARM_DESC(mode, "Mode of operation : 0 for round robin, 1 for active-backup, 2 for xor");
|
548 |
|
|
MODULE_PARM(primary, "s");
|
549 |
|
|
MODULE_PARM_DESC(primary, "Primary network device to use");
|
550 |
|
|
MODULE_PARM(lacp_rate, "s");
|
551 |
|
|
MODULE_PARM_DESC(lacp_rate, "LACPDU tx rate to request from 802.3ad partner (slow/fast)");
|
552 |
|
|
MODULE_PARM(arp_interval, "i");
|
553 |
|
|
MODULE_PARM_DESC(arp_interval, "arp interval in milliseconds");
|
554 |
|
|
MODULE_PARM(arp_ip_target, "1-" __MODULE_STRING(BOND_MAX_ARP_TARGETS) "s");
|
555 |
|
|
MODULE_PARM_DESC(arp_ip_target, "arp targets in n.n.n.n form");
|
556 |
|
|
|
557 |
|
|
/*----------------------------- Global variables ----------------------------*/
|
558 |
|
|
|
559 |
|
|
static const char *version =
|
560 |
|
|
DRV_DESCRIPTION ": v" DRV_VERSION " (" DRV_RELDATE ")\n";
|
561 |
|
|
|
562 |
|
|
static LIST_HEAD(bond_dev_list);
|
563 |
|
|
|
564 |
|
|
#ifdef CONFIG_PROC_FS
|
565 |
|
|
static struct proc_dir_entry *bond_proc_dir = NULL;
|
566 |
|
|
#endif
|
567 |
|
|
|
568 |
|
|
static u32 arp_target[BOND_MAX_ARP_TARGETS] = { 0, } ;
|
569 |
|
|
static int arp_ip_count = 0;
|
570 |
|
|
static u32 my_ip = 0;
|
571 |
|
|
static int bond_mode = BOND_MODE_ROUNDROBIN;
|
572 |
|
|
static int lacp_fast = 0;
|
573 |
|
|
static int app_abi_ver = 0;
|
574 |
|
|
static int orig_app_abi_ver = -1; /* This is used to save the first ABI version
|
575 |
|
|
* we receive from the application. Once set,
|
576 |
|
|
* it won't be changed, and the module will
|
577 |
|
|
* refuse to enslave/release interfaces if the
|
578 |
|
|
* command comes from an application using
|
579 |
|
|
* another ABI version.
|
580 |
|
|
*/
|
581 |
|
|
|
582 |
|
|
struct bond_parm_tbl {
|
583 |
|
|
char *modename;
|
584 |
|
|
int mode;
|
585 |
|
|
};
|
586 |
|
|
|
587 |
|
|
static struct bond_parm_tbl bond_lacp_tbl[] = {
|
588 |
|
|
{ "slow", AD_LACP_SLOW},
|
589 |
|
|
{ "fast", AD_LACP_FAST},
|
590 |
|
|
{ NULL, -1},
|
591 |
|
|
};
|
592 |
|
|
|
593 |
|
|
static struct bond_parm_tbl bond_mode_tbl[] = {
|
594 |
|
|
{ "balance-rr", BOND_MODE_ROUNDROBIN},
|
595 |
|
|
{ "active-backup", BOND_MODE_ACTIVEBACKUP},
|
596 |
|
|
{ "balance-xor", BOND_MODE_XOR},
|
597 |
|
|
{ "broadcast", BOND_MODE_BROADCAST},
|
598 |
|
|
{ "802.3ad", BOND_MODE_8023AD},
|
599 |
|
|
{ "balance-tlb", BOND_MODE_TLB},
|
600 |
|
|
{ "balance-alb", BOND_MODE_ALB},
|
601 |
|
|
{ NULL, -1},
|
602 |
|
|
};
|
603 |
|
|
|
604 |
|
|
/*-------------------------- Forward declarations ---------------------------*/
|
605 |
|
|
|
606 |
|
|
static inline void bond_set_mode_ops(struct net_device *bond_dev, int mode);
|
607 |
|
|
|
608 |
|
|
/*---------------------------- General routines -----------------------------*/
|
609 |
|
|
|
610 |
|
|
static const char *bond_mode_name(int mode)
|
611 |
|
|
{
|
612 |
|
|
switch (mode) {
|
613 |
|
|
case BOND_MODE_ROUNDROBIN :
|
614 |
|
|
return "load balancing (round-robin)";
|
615 |
|
|
case BOND_MODE_ACTIVEBACKUP :
|
616 |
|
|
return "fault-tolerance (active-backup)";
|
617 |
|
|
case BOND_MODE_XOR :
|
618 |
|
|
return "load balancing (xor)";
|
619 |
|
|
case BOND_MODE_BROADCAST :
|
620 |
|
|
return "fault-tolerance (broadcast)";
|
621 |
|
|
case BOND_MODE_8023AD:
|
622 |
|
|
return "IEEE 802.3ad Dynamic link aggregation";
|
623 |
|
|
case BOND_MODE_TLB:
|
624 |
|
|
return "transmit load balancing";
|
625 |
|
|
case BOND_MODE_ALB:
|
626 |
|
|
return "adaptive load balancing";
|
627 |
|
|
default:
|
628 |
|
|
return "unknown";
|
629 |
|
|
}
|
630 |
|
|
}
|
631 |
|
|
|
632 |
|
|
/*---------------------------------- VLAN -----------------------------------*/
|
633 |
|
|
|
634 |
|
|
/**
|
635 |
|
|
* bond_add_vlan - add a new vlan id on bond
|
636 |
|
|
* @bond: bond that got the notification
|
637 |
|
|
* @vlan_id: the vlan id to add
|
638 |
|
|
*
|
639 |
|
|
* Returns -ENOMEM if allocation failed.
|
640 |
|
|
*/
|
641 |
|
|
static int bond_add_vlan(struct bonding *bond, unsigned short vlan_id)
|
642 |
|
|
{
|
643 |
|
|
struct vlan_entry *vlan;
|
644 |
|
|
|
645 |
|
|
dprintk("bond: %s, vlan id %d\n",
|
646 |
|
|
(bond ? bond->dev->name: "None"), vlan_id);
|
647 |
|
|
|
648 |
|
|
vlan = kmalloc(sizeof(struct vlan_entry), GFP_KERNEL);
|
649 |
|
|
if (!vlan) {
|
650 |
|
|
return -ENOMEM;
|
651 |
|
|
}
|
652 |
|
|
|
653 |
|
|
INIT_LIST_HEAD(&vlan->vlan_list);
|
654 |
|
|
vlan->vlan_id = vlan_id;
|
655 |
|
|
|
656 |
|
|
write_lock_bh(&bond->lock);
|
657 |
|
|
|
658 |
|
|
list_add_tail(&vlan->vlan_list, &bond->vlan_list);
|
659 |
|
|
|
660 |
|
|
write_unlock_bh(&bond->lock);
|
661 |
|
|
|
662 |
|
|
dprintk("added VLAN ID %d on bond %s\n", vlan_id, bond->dev->name);
|
663 |
|
|
|
664 |
|
|
return 0;
|
665 |
|
|
}
|
666 |
|
|
|
667 |
|
|
/**
|
668 |
|
|
* bond_del_vlan - delete a vlan id from bond
|
669 |
|
|
* @bond: bond that got the notification
|
670 |
|
|
* @vlan_id: the vlan id to delete
|
671 |
|
|
*
|
672 |
|
|
* returns -ENODEV if @vlan_id was not found in @bond.
|
673 |
|
|
*/
|
674 |
|
|
static int bond_del_vlan(struct bonding *bond, unsigned short vlan_id)
|
675 |
|
|
{
|
676 |
|
|
struct vlan_entry *vlan, *next;
|
677 |
|
|
int res = -ENODEV;
|
678 |
|
|
|
679 |
|
|
dprintk("bond: %s, vlan id %d\n", bond->dev->name, vlan_id);
|
680 |
|
|
|
681 |
|
|
write_lock_bh(&bond->lock);
|
682 |
|
|
|
683 |
|
|
list_for_each_entry_safe(vlan, next, &bond->vlan_list, vlan_list) {
|
684 |
|
|
if (vlan->vlan_id == vlan_id) {
|
685 |
|
|
list_del(&vlan->vlan_list);
|
686 |
|
|
|
687 |
|
|
if ((bond->params.mode == BOND_MODE_TLB) ||
|
688 |
|
|
(bond->params.mode == BOND_MODE_ALB)) {
|
689 |
|
|
bond_alb_clear_vlan(bond, vlan_id);
|
690 |
|
|
}
|
691 |
|
|
|
692 |
|
|
dprintk("removed VLAN ID %d from bond %s\n", vlan_id,
|
693 |
|
|
bond->dev->name);
|
694 |
|
|
|
695 |
|
|
kfree(vlan);
|
696 |
|
|
|
697 |
|
|
if (list_empty(&bond->vlan_list) &&
|
698 |
|
|
(bond->slave_cnt == 0)) {
|
699 |
|
|
/* Last VLAN removed and no slaves, so
|
700 |
|
|
* restore block on adding VLANs. This will
|
701 |
|
|
* be removed once new slaves that are not
|
702 |
|
|
* VLAN challenged will be added.
|
703 |
|
|
*/
|
704 |
|
|
bond->dev->features |= NETIF_F_VLAN_CHALLENGED;
|
705 |
|
|
}
|
706 |
|
|
|
707 |
|
|
res = 0;
|
708 |
|
|
goto out;
|
709 |
|
|
}
|
710 |
|
|
}
|
711 |
|
|
|
712 |
|
|
dprintk("couldn't find VLAN ID %d in bond %s\n", vlan_id,
|
713 |
|
|
bond->dev->name);
|
714 |
|
|
|
715 |
|
|
out:
|
716 |
|
|
write_unlock_bh(&bond->lock);
|
717 |
|
|
return res;
|
718 |
|
|
}
|
719 |
|
|
|
720 |
|
|
/**
|
721 |
|
|
* bond_has_challenged_slaves
|
722 |
|
|
* @bond: the bond we're working on
|
723 |
|
|
*
|
724 |
|
|
* Searches the slave list. Returns 1 if a vlan challenged slave
|
725 |
|
|
* was found, 0 otherwise.
|
726 |
|
|
*
|
727 |
|
|
* Assumes bond->lock is held.
|
728 |
|
|
*/
|
729 |
|
|
static int bond_has_challenged_slaves(struct bonding *bond)
|
730 |
|
|
{
|
731 |
|
|
struct slave *slave;
|
732 |
|
|
int i;
|
733 |
|
|
|
734 |
|
|
bond_for_each_slave(bond, slave, i) {
|
735 |
|
|
if (slave->dev->features & NETIF_F_VLAN_CHALLENGED) {
|
736 |
|
|
dprintk("found VLAN challenged slave - %s\n",
|
737 |
|
|
slave->dev->name);
|
738 |
|
|
return 1;
|
739 |
|
|
}
|
740 |
|
|
}
|
741 |
|
|
|
742 |
|
|
dprintk("no VLAN challenged slaves found\n");
|
743 |
|
|
return 0;
|
744 |
|
|
}
|
745 |
|
|
|
746 |
|
|
/**
|
747 |
|
|
* bond_next_vlan - safely skip to the next item in the vlans list.
|
748 |
|
|
* @bond: the bond we're working on
|
749 |
|
|
* @curr: item we're advancing from
|
750 |
|
|
*
|
751 |
|
|
* Returns %NULL if list is empty, bond->next_vlan if @curr is %NULL,
|
752 |
|
|
* or @curr->next otherwise (even if it is @curr itself again).
|
753 |
|
|
*
|
754 |
|
|
* Caller must hold bond->lock
|
755 |
|
|
*/
|
756 |
|
|
struct vlan_entry *bond_next_vlan(struct bonding *bond, struct vlan_entry *curr)
|
757 |
|
|
{
|
758 |
|
|
struct vlan_entry *next, *last;
|
759 |
|
|
|
760 |
|
|
if (list_empty(&bond->vlan_list)) {
|
761 |
|
|
return NULL;
|
762 |
|
|
}
|
763 |
|
|
|
764 |
|
|
if (!curr) {
|
765 |
|
|
next = list_entry(bond->vlan_list.next,
|
766 |
|
|
struct vlan_entry, vlan_list);
|
767 |
|
|
} else {
|
768 |
|
|
last = list_entry(bond->vlan_list.prev,
|
769 |
|
|
struct vlan_entry, vlan_list);
|
770 |
|
|
if (last == curr) {
|
771 |
|
|
next = list_entry(bond->vlan_list.next,
|
772 |
|
|
struct vlan_entry, vlan_list);
|
773 |
|
|
} else {
|
774 |
|
|
next = list_entry(curr->vlan_list.next,
|
775 |
|
|
struct vlan_entry, vlan_list);
|
776 |
|
|
}
|
777 |
|
|
}
|
778 |
|
|
|
779 |
|
|
return next;
|
780 |
|
|
}
|
781 |
|
|
|
782 |
|
|
/**
|
783 |
|
|
* bond_dev_queue_xmit - Prepare skb for xmit.
|
784 |
|
|
*
|
785 |
|
|
* @bond: bond device that got this skb for tx.
|
786 |
|
|
* @skb: hw accel VLAN tagged skb to transmit
|
787 |
|
|
* @slave_dev: slave that is supposed to xmit this skbuff
|
788 |
|
|
*
|
789 |
|
|
* When the bond gets an skb to tarnsmit that is
|
790 |
|
|
* already hardware accelerated VLAN tagged, and it
|
791 |
|
|
* needs to relay this skb to a slave that is not
|
792 |
|
|
* hw accel capable, the skb needs to be "unaccelerated",
|
793 |
|
|
* i.e. strip the hwaccel tag and re-insert it as part
|
794 |
|
|
* of the payload.
|
795 |
|
|
*
|
796 |
|
|
* Assumption - once a VLAN device is created over the bond device, all
|
797 |
|
|
* packets are going to be hardware accelerated VLAN tagged since the IP
|
798 |
|
|
* binding is done over the VLAN device
|
799 |
|
|
*/
|
800 |
|
|
int bond_dev_queue_xmit(struct bonding *bond, struct sk_buff *skb, struct net_device *slave_dev)
|
801 |
|
|
{
|
802 |
|
|
unsigned short vlan_id;
|
803 |
|
|
int res;
|
804 |
|
|
|
805 |
|
|
if (!list_empty(&bond->vlan_list) &&
|
806 |
|
|
!(slave_dev->features & NETIF_F_HW_VLAN_TX)) {
|
807 |
|
|
res = vlan_get_tag(skb, &vlan_id);
|
808 |
|
|
if (res) {
|
809 |
|
|
return -EINVAL;
|
810 |
|
|
}
|
811 |
|
|
|
812 |
|
|
skb->dev = slave_dev;
|
813 |
|
|
skb = vlan_put_tag(skb, vlan_id);
|
814 |
|
|
if (!skb) {
|
815 |
|
|
/* vlan_put_tag() frees the skb in case of error,
|
816 |
|
|
* so return success here so the calling functions
|
817 |
|
|
* won't attempt to free is again.
|
818 |
|
|
*/
|
819 |
|
|
return 0;
|
820 |
|
|
}
|
821 |
|
|
} else {
|
822 |
|
|
skb->dev = slave_dev;
|
823 |
|
|
}
|
824 |
|
|
|
825 |
|
|
skb->priority = 1;
|
826 |
|
|
dev_queue_xmit(skb);
|
827 |
|
|
|
828 |
|
|
return 0;
|
829 |
|
|
}
|
830 |
|
|
|
831 |
|
|
/*
|
832 |
|
|
* In the following 3 functions, bond_vlan_rx_register(), bond_vlan_rx_add_vid
|
833 |
|
|
* and bond_vlan_rx_kill_vid, We don't protect the slave list iteration with a
|
834 |
|
|
* lock because:
|
835 |
|
|
* a. This operation is performed in IOCTL context,
|
836 |
|
|
* b. The operation is protected by the RTNL semaphore in the 8021q code,
|
837 |
|
|
* c. Holding a lock with BH disabled while directly calling a base driver
|
838 |
|
|
* entry point is generally a BAD idea.
|
839 |
|
|
*
|
840 |
|
|
* The design of synchronization/protection for this operation in the 8021q
|
841 |
|
|
* module is good for one or more VLAN devices over a single physical device
|
842 |
|
|
* and cannot be extended for a teaming solution like bonding, so there is a
|
843 |
|
|
* potential race condition here where a net device from the vlan group might
|
844 |
|
|
* be referenced (either by a base driver or the 8021q code) while it is being
|
845 |
|
|
* removed from the system. However, it turns out we're not making matters
|
846 |
|
|
* worse, and if it works for regular VLAN usage it will work here too.
|
847 |
|
|
*/
|
848 |
|
|
|
849 |
|
|
/**
|
850 |
|
|
* bond_vlan_rx_register - Propagates registration to slaves
|
851 |
|
|
* @bond_dev: bonding net device that got called
|
852 |
|
|
* @grp: vlan group being registered
|
853 |
|
|
*/
|
854 |
|
|
static void bond_vlan_rx_register(struct net_device *bond_dev, struct vlan_group *grp)
|
855 |
|
|
{
|
856 |
|
|
struct bonding *bond = bond_dev->priv;
|
857 |
|
|
struct slave *slave;
|
858 |
|
|
int i;
|
859 |
|
|
|
860 |
|
|
bond->vlgrp = grp;
|
861 |
|
|
|
862 |
|
|
bond_for_each_slave(bond, slave, i) {
|
863 |
|
|
struct net_device *slave_dev = slave->dev;
|
864 |
|
|
|
865 |
|
|
if ((slave_dev->features & NETIF_F_HW_VLAN_RX) &&
|
866 |
|
|
slave_dev->vlan_rx_register) {
|
867 |
|
|
slave_dev->vlan_rx_register(slave_dev, grp);
|
868 |
|
|
}
|
869 |
|
|
}
|
870 |
|
|
}
|
871 |
|
|
|
872 |
|
|
/**
|
873 |
|
|
* bond_vlan_rx_add_vid - Propagates adding an id to slaves
|
874 |
|
|
* @bond_dev: bonding net device that got called
|
875 |
|
|
* @vid: vlan id being added
|
876 |
|
|
*/
|
877 |
|
|
static void bond_vlan_rx_add_vid(struct net_device *bond_dev, uint16_t vid)
|
878 |
|
|
{
|
879 |
|
|
struct bonding *bond = bond_dev->priv;
|
880 |
|
|
struct slave *slave;
|
881 |
|
|
int i, res;
|
882 |
|
|
|
883 |
|
|
bond_for_each_slave(bond, slave, i) {
|
884 |
|
|
struct net_device *slave_dev = slave->dev;
|
885 |
|
|
|
886 |
|
|
if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) &&
|
887 |
|
|
slave_dev->vlan_rx_add_vid) {
|
888 |
|
|
slave_dev->vlan_rx_add_vid(slave_dev, vid);
|
889 |
|
|
}
|
890 |
|
|
}
|
891 |
|
|
|
892 |
|
|
res = bond_add_vlan(bond, vid);
|
893 |
|
|
if (res) {
|
894 |
|
|
printk(KERN_ERR DRV_NAME
|
895 |
|
|
": %s: Failed to add vlan id %d\n",
|
896 |
|
|
bond_dev->name, vid);
|
897 |
|
|
}
|
898 |
|
|
}
|
899 |
|
|
|
900 |
|
|
/**
|
901 |
|
|
* bond_vlan_rx_kill_vid - Propagates deleting an id to slaves
|
902 |
|
|
* @bond_dev: bonding net device that got called
|
903 |
|
|
* @vid: vlan id being removed
|
904 |
|
|
*/
|
905 |
|
|
static void bond_vlan_rx_kill_vid(struct net_device *bond_dev, uint16_t vid)
|
906 |
|
|
{
|
907 |
|
|
struct bonding *bond = bond_dev->priv;
|
908 |
|
|
struct slave *slave;
|
909 |
|
|
struct net_device *vlan_dev;
|
910 |
|
|
int i, res;
|
911 |
|
|
|
912 |
|
|
bond_for_each_slave(bond, slave, i) {
|
913 |
|
|
struct net_device *slave_dev = slave->dev;
|
914 |
|
|
|
915 |
|
|
if ((slave_dev->features & NETIF_F_HW_VLAN_FILTER) &&
|
916 |
|
|
slave_dev->vlan_rx_kill_vid) {
|
917 |
|
|
/* Save and then restore vlan_dev in the grp array,
|
918 |
|
|
* since the slave's driver might clear it.
|
919 |
|
|
*/
|
920 |
|
|
vlan_dev = bond->vlgrp->vlan_devices[vid];
|
921 |
|
|
slave_dev->vlan_rx_kill_vid(slave_dev, vid);
|
922 |
|
|
bond->vlgrp->vlan_devices[vid] = vlan_dev;
|
923 |
|
|
}
|
924 |
|
|
}
|
925 |
|
|
|
926 |
|
|
res = bond_del_vlan(bond, vid);
|
927 |
|
|
if (res) {
|
928 |
|
|
printk(KERN_ERR DRV_NAME
|
929 |
|
|
": %s: Failed to remove vlan id %d\n",
|
930 |
|
|
bond_dev->name, vid);
|
931 |
|
|
}
|
932 |
|
|
}
|
933 |
|
|
|
934 |
|
|
static void bond_add_vlans_on_slave(struct bonding *bond, struct net_device *slave_dev)
|
935 |
|
|
{
|
936 |
|
|
struct vlan_entry *vlan;
|
937 |
|
|
|
938 |
|
|
write_lock_bh(&bond->lock);
|
939 |
|
|
|
940 |
|
|
if (list_empty(&bond->vlan_list)) {
|
941 |
|
|
goto out;
|
942 |
|
|
}
|
943 |
|
|
|
944 |
|
|
if ((slave_dev->features & NETIF_F_HW_VLAN_RX) &&
|
945 |
|
|
slave_dev->vlan_rx_register) {
|
946 |
|
|
slave_dev->vlan_rx_register(slave_dev, bond->vlgrp);
|
947 |
|
|
}
|
948 |
|
|
|
949 |
|
|
if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) ||
|
950 |
|
|
!(slave_dev->vlan_rx_add_vid)) {
|
951 |
|
|
goto out;
|
952 |
|
|
}
|
953 |
|
|
|
954 |
|
|
list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
|
955 |
|
|
slave_dev->vlan_rx_add_vid(slave_dev, vlan->vlan_id);
|
956 |
|
|
}
|
957 |
|
|
|
958 |
|
|
out:
|
959 |
|
|
write_unlock_bh(&bond->lock);
|
960 |
|
|
}
|
961 |
|
|
|
962 |
|
|
static void bond_del_vlans_from_slave(struct bonding *bond, struct net_device *slave_dev)
|
963 |
|
|
{
|
964 |
|
|
struct vlan_entry *vlan;
|
965 |
|
|
struct net_device *vlan_dev;
|
966 |
|
|
|
967 |
|
|
write_lock_bh(&bond->lock);
|
968 |
|
|
|
969 |
|
|
if (list_empty(&bond->vlan_list)) {
|
970 |
|
|
goto out;
|
971 |
|
|
}
|
972 |
|
|
|
973 |
|
|
if (!(slave_dev->features & NETIF_F_HW_VLAN_FILTER) ||
|
974 |
|
|
!(slave_dev->vlan_rx_kill_vid)) {
|
975 |
|
|
goto unreg;
|
976 |
|
|
}
|
977 |
|
|
|
978 |
|
|
list_for_each_entry(vlan, &bond->vlan_list, vlan_list) {
|
979 |
|
|
/* Save and then restore vlan_dev in the grp array,
|
980 |
|
|
* since the slave's driver might clear it.
|
981 |
|
|
*/
|
982 |
|
|
vlan_dev = bond->vlgrp->vlan_devices[vlan->vlan_id];
|
983 |
|
|
slave_dev->vlan_rx_kill_vid(slave_dev, vlan->vlan_id);
|
984 |
|
|
bond->vlgrp->vlan_devices[vlan->vlan_id] = vlan_dev;
|
985 |
|
|
}
|
986 |
|
|
|
987 |
|
|
unreg:
|
988 |
|
|
if ((slave_dev->features & NETIF_F_HW_VLAN_RX) &&
|
989 |
|
|
slave_dev->vlan_rx_register) {
|
990 |
|
|
slave_dev->vlan_rx_register(slave_dev, NULL);
|
991 |
|
|
}
|
992 |
|
|
|
993 |
|
|
out:
|
994 |
|
|
write_unlock_bh(&bond->lock);
|
995 |
|
|
}
|
996 |
|
|
|
997 |
|
|
/*------------------------------- Link status -------------------------------*/
|
998 |
|
|
|
999 |
|
|
/*
|
1000 |
|
|
* Get link speed and duplex from the slave's base driver
|
1001 |
|
|
* using ethtool. If for some reason the call fails or the
|
1002 |
|
|
* values are invalid, fake speed and duplex to 100/Full
|
1003 |
|
|
* and return error.
|
1004 |
|
|
*/
|
1005 |
|
|
static int bond_update_speed_duplex(struct slave *slave)
|
1006 |
|
|
{
|
1007 |
|
|
struct net_device *slave_dev = slave->dev;
|
1008 |
|
|
static int (* ioctl)(struct net_device *, struct ifreq *, int);
|
1009 |
|
|
struct ifreq ifr;
|
1010 |
|
|
struct ethtool_cmd etool;
|
1011 |
|
|
|
1012 |
|
|
/* Fake speed and duplex */
|
1013 |
|
|
slave->speed = SPEED_100;
|
1014 |
|
|
slave->duplex = DUPLEX_FULL;
|
1015 |
|
|
|
1016 |
|
|
if (slave_dev->ethtool_ops) {
|
1017 |
|
|
u32 res;
|
1018 |
|
|
|
1019 |
|
|
if (!slave_dev->ethtool_ops->get_settings) {
|
1020 |
|
|
return -1;
|
1021 |
|
|
}
|
1022 |
|
|
|
1023 |
|
|
res = slave_dev->ethtool_ops->get_settings(slave_dev, &etool);
|
1024 |
|
|
if (res < 0) {
|
1025 |
|
|
return -1;
|
1026 |
|
|
}
|
1027 |
|
|
|
1028 |
|
|
goto verify;
|
1029 |
|
|
}
|
1030 |
|
|
|
1031 |
|
|
ioctl = slave_dev->do_ioctl;
|
1032 |
|
|
strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ);
|
1033 |
|
|
etool.cmd = ETHTOOL_GSET;
|
1034 |
|
|
ifr.ifr_data = (char*)&etool;
|
1035 |
|
|
if (!ioctl || (IOCTL(slave_dev, &ifr, SIOCETHTOOL) < 0)) {
|
1036 |
|
|
return -1;
|
1037 |
|
|
}
|
1038 |
|
|
|
1039 |
|
|
verify:
|
1040 |
|
|
switch (etool.speed) {
|
1041 |
|
|
case SPEED_10:
|
1042 |
|
|
case SPEED_100:
|
1043 |
|
|
case SPEED_1000:
|
1044 |
|
|
break;
|
1045 |
|
|
default:
|
1046 |
|
|
return -1;
|
1047 |
|
|
}
|
1048 |
|
|
|
1049 |
|
|
switch (etool.duplex) {
|
1050 |
|
|
case DUPLEX_FULL:
|
1051 |
|
|
case DUPLEX_HALF:
|
1052 |
|
|
break;
|
1053 |
|
|
default:
|
1054 |
|
|
return -1;
|
1055 |
|
|
}
|
1056 |
|
|
|
1057 |
|
|
slave->speed = etool.speed;
|
1058 |
|
|
slave->duplex = etool.duplex;
|
1059 |
|
|
|
1060 |
|
|
return 0;
|
1061 |
|
|
}
|
1062 |
|
|
|
1063 |
|
|
/*
|
1064 |
|
|
* if <dev> supports MII link status reporting, check its link status.
|
1065 |
|
|
*
|
1066 |
|
|
* We either do MII/ETHTOOL ioctls, or check netif_carrier_ok(),
|
1067 |
|
|
* depening upon the setting of the use_carrier parameter.
|
1068 |
|
|
*
|
1069 |
|
|
* Return either BMSR_LSTATUS, meaning that the link is up (or we
|
1070 |
|
|
* can't tell and just pretend it is), or 0, meaning that the link is
|
1071 |
|
|
* down.
|
1072 |
|
|
*
|
1073 |
|
|
* If reporting is non-zero, instead of faking link up, return -1 if
|
1074 |
|
|
* both ETHTOOL and MII ioctls fail (meaning the device does not
|
1075 |
|
|
* support them). If use_carrier is set, return whatever it says.
|
1076 |
|
|
* It'd be nice if there was a good way to tell if a driver supports
|
1077 |
|
|
* netif_carrier, but there really isn't.
|
1078 |
|
|
*/
|
1079 |
|
|
static int bond_check_dev_link(struct bonding *bond, struct net_device *slave_dev, int reporting)
|
1080 |
|
|
{
|
1081 |
|
|
static int (* ioctl)(struct net_device *, struct ifreq *, int);
|
1082 |
|
|
struct ifreq ifr;
|
1083 |
|
|
struct mii_ioctl_data *mii;
|
1084 |
|
|
struct ethtool_value etool;
|
1085 |
|
|
|
1086 |
|
|
if (bond->params.use_carrier) {
|
1087 |
|
|
return netif_carrier_ok(slave_dev) ? BMSR_LSTATUS : 0;
|
1088 |
|
|
}
|
1089 |
|
|
|
1090 |
|
|
ioctl = slave_dev->do_ioctl;
|
1091 |
|
|
if (ioctl) {
|
1092 |
|
|
/* TODO: set pointer to correct ioctl on a per team member */
|
1093 |
|
|
/* bases to make this more efficient. that is, once */
|
1094 |
|
|
/* we determine the correct ioctl, we will always */
|
1095 |
|
|
/* call it and not the others for that team */
|
1096 |
|
|
/* member. */
|
1097 |
|
|
|
1098 |
|
|
/*
|
1099 |
|
|
* We cannot assume that SIOCGMIIPHY will also read a
|
1100 |
|
|
* register; not all network drivers (e.g., e100)
|
1101 |
|
|
* support that.
|
1102 |
|
|
*/
|
1103 |
|
|
|
1104 |
|
|
/* Yes, the mii is overlaid on the ifreq.ifr_ifru */
|
1105 |
|
|
strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ);
|
1106 |
|
|
mii = (struct mii_ioctl_data *)&ifr.ifr_data;
|
1107 |
|
|
if (IOCTL(slave_dev, &ifr, SIOCGMIIPHY) == 0) {
|
1108 |
|
|
mii->reg_num = MII_BMSR;
|
1109 |
|
|
if (IOCTL(slave_dev, &ifr, SIOCGMIIREG) == 0) {
|
1110 |
|
|
return (mii->val_out & BMSR_LSTATUS);
|
1111 |
|
|
}
|
1112 |
|
|
}
|
1113 |
|
|
}
|
1114 |
|
|
|
1115 |
|
|
/* try SIOCETHTOOL ioctl, some drivers cache ETHTOOL_GLINK */
|
1116 |
|
|
/* for a period of time so we attempt to get link status */
|
1117 |
|
|
/* from it last if the above MII ioctls fail... */
|
1118 |
|
|
if (slave_dev->ethtool_ops) {
|
1119 |
|
|
if (slave_dev->ethtool_ops->get_link) {
|
1120 |
|
|
u32 link;
|
1121 |
|
|
|
1122 |
|
|
link = slave_dev->ethtool_ops->get_link(slave_dev);
|
1123 |
|
|
|
1124 |
|
|
return link ? BMSR_LSTATUS : 0;
|
1125 |
|
|
}
|
1126 |
|
|
}
|
1127 |
|
|
|
1128 |
|
|
if (ioctl) {
|
1129 |
|
|
strncpy(ifr.ifr_name, slave_dev->name, IFNAMSIZ);
|
1130 |
|
|
etool.cmd = ETHTOOL_GLINK;
|
1131 |
|
|
ifr.ifr_data = (char*)&etool;
|
1132 |
|
|
if (IOCTL(slave_dev, &ifr, SIOCETHTOOL) == 0) {
|
1133 |
|
|
if (etool.data == 1) {
|
1134 |
|
|
return BMSR_LSTATUS;
|
1135 |
|
|
} else {
|
1136 |
|
|
dprintk("SIOCETHTOOL shows link down\n");
|
1137 |
|
|
return 0;
|
1138 |
|
|
}
|
1139 |
|
|
}
|
1140 |
|
|
}
|
1141 |
|
|
|
1142 |
|
|
/*
|
1143 |
|
|
* If reporting, report that either there's no dev->do_ioctl,
|
1144 |
|
|
* or both SIOCGMIIREG and SIOCETHTOOL failed (meaning that we
|
1145 |
|
|
* cannot report link status). If not reporting, pretend
|
1146 |
|
|
* we're ok.
|
1147 |
|
|
*/
|
1148 |
|
|
return (reporting ? -1 : BMSR_LSTATUS);
|
1149 |
|
|
}
|
1150 |
|
|
|
1151 |
|
|
/*----------------------------- Multicast list ------------------------------*/
|
1152 |
|
|
|
1153 |
|
|
/*
|
1154 |
|
|
* Returns 0 if dmi1 and dmi2 are the same, non-0 otherwise
|
1155 |
|
|
*/
|
1156 |
|
|
static inline int bond_is_dmi_same(struct dev_mc_list *dmi1, struct dev_mc_list *dmi2)
|
1157 |
|
|
{
|
1158 |
|
|
return memcmp(dmi1->dmi_addr, dmi2->dmi_addr, dmi1->dmi_addrlen) == 0 &&
|
1159 |
|
|
dmi1->dmi_addrlen == dmi2->dmi_addrlen;
|
1160 |
|
|
}
|
1161 |
|
|
|
1162 |
|
|
/*
|
1163 |
|
|
* returns dmi entry if found, NULL otherwise
|
1164 |
|
|
*/
|
1165 |
|
|
static struct dev_mc_list *bond_mc_list_find_dmi(struct dev_mc_list *dmi, struct dev_mc_list *mc_list)
|
1166 |
|
|
{
|
1167 |
|
|
struct dev_mc_list *idmi;
|
1168 |
|
|
|
1169 |
|
|
for (idmi = mc_list; idmi; idmi = idmi->next) {
|
1170 |
|
|
if (bond_is_dmi_same(dmi, idmi)) {
|
1171 |
|
|
return idmi;
|
1172 |
|
|
}
|
1173 |
|
|
}
|
1174 |
|
|
|
1175 |
|
|
return NULL;
|
1176 |
|
|
}
|
1177 |
|
|
|
1178 |
|
|
/*
|
1179 |
|
|
* Push the promiscuity flag down to appropriate slaves
|
1180 |
|
|
*/
|
1181 |
|
|
static void bond_set_promiscuity(struct bonding *bond, int inc)
|
1182 |
|
|
{
|
1183 |
|
|
if (USES_PRIMARY(bond->params.mode)) {
|
1184 |
|
|
/* write lock already acquired */
|
1185 |
|
|
if (bond->curr_active_slave) {
|
1186 |
|
|
dev_set_promiscuity(bond->curr_active_slave->dev, inc);
|
1187 |
|
|
}
|
1188 |
|
|
} else {
|
1189 |
|
|
struct slave *slave;
|
1190 |
|
|
int i;
|
1191 |
|
|
bond_for_each_slave(bond, slave, i) {
|
1192 |
|
|
dev_set_promiscuity(slave->dev, inc);
|
1193 |
|
|
}
|
1194 |
|
|
}
|
1195 |
|
|
}
|
1196 |
|
|
|
1197 |
|
|
/*
|
1198 |
|
|
* Push the allmulti flag down to all slaves
|
1199 |
|
|
*/
|
1200 |
|
|
static void bond_set_allmulti(struct bonding *bond, int inc)
|
1201 |
|
|
{
|
1202 |
|
|
if (USES_PRIMARY(bond->params.mode)) {
|
1203 |
|
|
/* write lock already acquired */
|
1204 |
|
|
if (bond->curr_active_slave) {
|
1205 |
|
|
dev_set_allmulti(bond->curr_active_slave->dev, inc);
|
1206 |
|
|
}
|
1207 |
|
|
} else {
|
1208 |
|
|
struct slave *slave;
|
1209 |
|
|
int i;
|
1210 |
|
|
bond_for_each_slave(bond, slave, i) {
|
1211 |
|
|
dev_set_allmulti(slave->dev, inc);
|
1212 |
|
|
}
|
1213 |
|
|
}
|
1214 |
|
|
}
|
1215 |
|
|
|
1216 |
|
|
/*
|
1217 |
|
|
* Add a Multicast address to slaves
|
1218 |
|
|
* according to mode
|
1219 |
|
|
*/
|
1220 |
|
|
static void bond_mc_add(struct bonding *bond, void *addr, int alen)
|
1221 |
|
|
{
|
1222 |
|
|
if (USES_PRIMARY(bond->params.mode)) {
|
1223 |
|
|
/* write lock already acquired */
|
1224 |
|
|
if (bond->curr_active_slave) {
|
1225 |
|
|
dev_mc_add(bond->curr_active_slave->dev, addr, alen, 0);
|
1226 |
|
|
}
|
1227 |
|
|
} else {
|
1228 |
|
|
struct slave *slave;
|
1229 |
|
|
int i;
|
1230 |
|
|
bond_for_each_slave(bond, slave, i) {
|
1231 |
|
|
dev_mc_add(slave->dev, addr, alen, 0);
|
1232 |
|
|
}
|
1233 |
|
|
}
|
1234 |
|
|
}
|
1235 |
|
|
|
1236 |
|
|
/*
|
1237 |
|
|
* Remove a multicast address from slave
|
1238 |
|
|
* according to mode
|
1239 |
|
|
*/
|
1240 |
|
|
static void bond_mc_delete(struct bonding *bond, void *addr, int alen)
|
1241 |
|
|
{
|
1242 |
|
|
if (USES_PRIMARY(bond->params.mode)) {
|
1243 |
|
|
/* write lock already acquired */
|
1244 |
|
|
if (bond->curr_active_slave) {
|
1245 |
|
|
dev_mc_delete(bond->curr_active_slave->dev, addr, alen, 0);
|
1246 |
|
|
}
|
1247 |
|
|
} else {
|
1248 |
|
|
struct slave *slave;
|
1249 |
|
|
int i;
|
1250 |
|
|
bond_for_each_slave(bond, slave, i) {
|
1251 |
|
|
dev_mc_delete(slave->dev, addr, alen, 0);
|
1252 |
|
|
}
|
1253 |
|
|
}
|
1254 |
|
|
}
|
1255 |
|
|
|
1256 |
|
|
/*
|
1257 |
|
|
* Totally destroys the mc_list in bond
|
1258 |
|
|
*/
|
1259 |
|
|
static void bond_mc_list_destroy(struct bonding *bond)
|
1260 |
|
|
{
|
1261 |
|
|
struct dev_mc_list *dmi;
|
1262 |
|
|
|
1263 |
|
|
dmi = bond->mc_list;
|
1264 |
|
|
while (dmi) {
|
1265 |
|
|
bond->mc_list = dmi->next;
|
1266 |
|
|
kfree(dmi);
|
1267 |
|
|
dmi = bond->mc_list;
|
1268 |
|
|
}
|
1269 |
|
|
}
|
1270 |
|
|
|
1271 |
|
|
/*
|
1272 |
|
|
* Copy all the Multicast addresses from src to the bonding device dst
|
1273 |
|
|
*/
|
1274 |
|
|
static int bond_mc_list_copy(struct dev_mc_list *mc_list, struct bonding *bond, int gpf_flag)
|
1275 |
|
|
{
|
1276 |
|
|
struct dev_mc_list *dmi, *new_dmi;
|
1277 |
|
|
|
1278 |
|
|
for (dmi = mc_list; dmi; dmi = dmi->next) {
|
1279 |
|
|
new_dmi = kmalloc(sizeof(struct dev_mc_list), gpf_flag);
|
1280 |
|
|
|
1281 |
|
|
if (!new_dmi) {
|
1282 |
|
|
/* FIXME: Potential memory leak !!! */
|
1283 |
|
|
return -ENOMEM;
|
1284 |
|
|
}
|
1285 |
|
|
|
1286 |
|
|
new_dmi->next = bond->mc_list;
|
1287 |
|
|
bond->mc_list = new_dmi;
|
1288 |
|
|
new_dmi->dmi_addrlen = dmi->dmi_addrlen;
|
1289 |
|
|
memcpy(new_dmi->dmi_addr, dmi->dmi_addr, dmi->dmi_addrlen);
|
1290 |
|
|
new_dmi->dmi_users = dmi->dmi_users;
|
1291 |
|
|
new_dmi->dmi_gusers = dmi->dmi_gusers;
|
1292 |
|
|
}
|
1293 |
|
|
|
1294 |
|
|
return 0;
|
1295 |
|
|
}
|
1296 |
|
|
|
1297 |
|
|
/*
|
1298 |
|
|
* flush all members of flush->mc_list from device dev->mc_list
|
1299 |
|
|
*/
|
1300 |
|
|
static void bond_mc_list_flush(struct net_device *bond_dev, struct net_device *slave_dev)
|
1301 |
|
|
{
|
1302 |
|
|
struct bonding *bond = bond_dev->priv;
|
1303 |
|
|
struct dev_mc_list *dmi;
|
1304 |
|
|
|
1305 |
|
|
for (dmi = bond_dev->mc_list; dmi; dmi = dmi->next) {
|
1306 |
|
|
dev_mc_delete(slave_dev, dmi->dmi_addr, dmi->dmi_addrlen, 0);
|
1307 |
|
|
}
|
1308 |
|
|
|
1309 |
|
|
if (bond->params.mode == BOND_MODE_8023AD) {
|
1310 |
|
|
/* del lacpdu mc addr from mc list */
|
1311 |
|
|
u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
|
1312 |
|
|
|
1313 |
|
|
dev_mc_delete(slave_dev, lacpdu_multicast, ETH_ALEN, 0);
|
1314 |
|
|
}
|
1315 |
|
|
}
|
1316 |
|
|
|
1317 |
|
|
/*--------------------------- Active slave change ---------------------------*/
|
1318 |
|
|
|
1319 |
|
|
/*
|
1320 |
|
|
* Update the mc list and multicast-related flags for the new and
|
1321 |
|
|
* old active slaves (if any) according to the multicast mode, and
|
1322 |
|
|
* promiscuous flags unconditionally.
|
1323 |
|
|
*/
|
1324 |
|
|
static void bond_mc_swap(struct bonding *bond, struct slave *new_active, struct slave *old_active)
|
1325 |
|
|
{
|
1326 |
|
|
struct dev_mc_list *dmi;
|
1327 |
|
|
|
1328 |
|
|
if (!USES_PRIMARY(bond->params.mode)) {
|
1329 |
|
|
/* nothing to do - mc list is already up-to-date on
|
1330 |
|
|
* all slaves
|
1331 |
|
|
*/
|
1332 |
|
|
return;
|
1333 |
|
|
}
|
1334 |
|
|
|
1335 |
|
|
if (old_active) {
|
1336 |
|
|
if (bond->dev->flags & IFF_PROMISC) {
|
1337 |
|
|
dev_set_promiscuity(old_active->dev, -1);
|
1338 |
|
|
}
|
1339 |
|
|
|
1340 |
|
|
if (bond->dev->flags & IFF_ALLMULTI) {
|
1341 |
|
|
dev_set_allmulti(old_active->dev, -1);
|
1342 |
|
|
}
|
1343 |
|
|
|
1344 |
|
|
for (dmi = bond->dev->mc_list; dmi; dmi = dmi->next) {
|
1345 |
|
|
dev_mc_delete(old_active->dev, dmi->dmi_addr, dmi->dmi_addrlen, 0);
|
1346 |
|
|
}
|
1347 |
|
|
}
|
1348 |
|
|
|
1349 |
|
|
if (new_active) {
|
1350 |
|
|
if (bond->dev->flags & IFF_PROMISC) {
|
1351 |
|
|
dev_set_promiscuity(new_active->dev, 1);
|
1352 |
|
|
}
|
1353 |
|
|
|
1354 |
|
|
if (bond->dev->flags & IFF_ALLMULTI) {
|
1355 |
|
|
dev_set_allmulti(new_active->dev, 1);
|
1356 |
|
|
}
|
1357 |
|
|
|
1358 |
|
|
for (dmi = bond->dev->mc_list; dmi; dmi = dmi->next) {
|
1359 |
|
|
dev_mc_add(new_active->dev, dmi->dmi_addr, dmi->dmi_addrlen, 0);
|
1360 |
|
|
}
|
1361 |
|
|
}
|
1362 |
|
|
}
|
1363 |
|
|
|
1364 |
|
|
/**
|
1365 |
|
|
* find_best_interface - select the best available slave to be the active one
|
1366 |
|
|
* @bond: our bonding struct
|
1367 |
|
|
*
|
1368 |
|
|
* Warning: Caller must hold curr_slave_lock for writing.
|
1369 |
|
|
*/
|
1370 |
|
|
static struct slave *bond_find_best_slave(struct bonding *bond)
|
1371 |
|
|
{
|
1372 |
|
|
struct slave *new_active, *old_active;
|
1373 |
|
|
struct slave *bestslave = NULL;
|
1374 |
|
|
int mintime = bond->params.updelay;
|
1375 |
|
|
int i;
|
1376 |
|
|
|
1377 |
|
|
new_active = old_active = bond->curr_active_slave;
|
1378 |
|
|
|
1379 |
|
|
if (!new_active) { /* there were no active slaves left */
|
1380 |
|
|
if (bond->slave_cnt > 0) { /* found one slave */
|
1381 |
|
|
new_active = bond->first_slave;
|
1382 |
|
|
} else {
|
1383 |
|
|
return NULL; /* still no slave, return NULL */
|
1384 |
|
|
}
|
1385 |
|
|
}
|
1386 |
|
|
|
1387 |
|
|
/* first try the primary link; if arping, a link must tx/rx traffic
|
1388 |
|
|
* before it can be considered the curr_active_slave - also, we would skip
|
1389 |
|
|
* slaves between the curr_active_slave and primary_slave that may be up
|
1390 |
|
|
* and able to arp
|
1391 |
|
|
*/
|
1392 |
|
|
if ((bond->primary_slave) &&
|
1393 |
|
|
(!bond->params.arp_interval) &&
|
1394 |
|
|
(IS_UP(bond->primary_slave->dev))) {
|
1395 |
|
|
new_active = bond->primary_slave;
|
1396 |
|
|
}
|
1397 |
|
|
|
1398 |
|
|
/* remember where to stop iterating over the slaves */
|
1399 |
|
|
old_active = new_active;
|
1400 |
|
|
|
1401 |
|
|
bond_for_each_slave_from(bond, new_active, i, old_active) {
|
1402 |
|
|
if (IS_UP(new_active->dev)) {
|
1403 |
|
|
if (new_active->link == BOND_LINK_UP) {
|
1404 |
|
|
return new_active;
|
1405 |
|
|
} else if (new_active->link == BOND_LINK_BACK) {
|
1406 |
|
|
/* link up, but waiting for stabilization */
|
1407 |
|
|
if (new_active->delay < mintime) {
|
1408 |
|
|
mintime = new_active->delay;
|
1409 |
|
|
bestslave = new_active;
|
1410 |
|
|
}
|
1411 |
|
|
}
|
1412 |
|
|
}
|
1413 |
|
|
}
|
1414 |
|
|
|
1415 |
|
|
return bestslave;
|
1416 |
|
|
}
|
1417 |
|
|
|
1418 |
|
|
/**
|
1419 |
|
|
* change_active_interface - change the active slave into the specified one
|
1420 |
|
|
* @bond: our bonding struct
|
1421 |
|
|
* @new: the new slave to make the active one
|
1422 |
|
|
*
|
1423 |
|
|
* Set the new slave to the bond's settings and unset them on the old
|
1424 |
|
|
* curr_active_slave.
|
1425 |
|
|
* Setting include flags, mc-list, promiscuity, allmulti, etc.
|
1426 |
|
|
*
|
1427 |
|
|
* If @new's link state is %BOND_LINK_BACK we'll set it to %BOND_LINK_UP,
|
1428 |
|
|
* because it is apparently the best available slave we have, even though its
|
1429 |
|
|
* updelay hasn't timed out yet.
|
1430 |
|
|
*
|
1431 |
|
|
* Warning: Caller must hold curr_slave_lock for writing.
|
1432 |
|
|
*/
|
1433 |
|
|
static void bond_change_active_slave(struct bonding *bond, struct slave *new_active)
|
1434 |
|
|
{
|
1435 |
|
|
struct slave *old_active = bond->curr_active_slave;
|
1436 |
|
|
|
1437 |
|
|
if (old_active == new_active) {
|
1438 |
|
|
return;
|
1439 |
|
|
}
|
1440 |
|
|
|
1441 |
|
|
if (new_active) {
|
1442 |
|
|
if (new_active->link == BOND_LINK_BACK) {
|
1443 |
|
|
if (USES_PRIMARY(bond->params.mode)) {
|
1444 |
|
|
printk(KERN_INFO DRV_NAME
|
1445 |
|
|
": %s: making interface %s the new "
|
1446 |
|
|
"active one %d ms earlier.\n",
|
1447 |
|
|
bond->dev->name, new_active->dev->name,
|
1448 |
|
|
(bond->params.updelay - new_active->delay) * bond->params.miimon);
|
1449 |
|
|
}
|
1450 |
|
|
|
1451 |
|
|
new_active->delay = 0;
|
1452 |
|
|
new_active->link = BOND_LINK_UP;
|
1453 |
|
|
new_active->jiffies = jiffies;
|
1454 |
|
|
|
1455 |
|
|
if (bond->params.mode == BOND_MODE_8023AD) {
|
1456 |
|
|
bond_3ad_handle_link_change(new_active, BOND_LINK_UP);
|
1457 |
|
|
}
|
1458 |
|
|
|
1459 |
|
|
if ((bond->params.mode == BOND_MODE_TLB) ||
|
1460 |
|
|
(bond->params.mode == BOND_MODE_ALB)) {
|
1461 |
|
|
bond_alb_handle_link_change(bond, new_active, BOND_LINK_UP);
|
1462 |
|
|
}
|
1463 |
|
|
} else {
|
1464 |
|
|
if (USES_PRIMARY(bond->params.mode)) {
|
1465 |
|
|
printk(KERN_INFO DRV_NAME
|
1466 |
|
|
": %s: making interface %s the new "
|
1467 |
|
|
"active one.\n",
|
1468 |
|
|
bond->dev->name, new_active->dev->name);
|
1469 |
|
|
}
|
1470 |
|
|
}
|
1471 |
|
|
}
|
1472 |
|
|
|
1473 |
|
|
if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) {
|
1474 |
|
|
if (old_active) {
|
1475 |
|
|
bond_set_slave_inactive_flags(old_active);
|
1476 |
|
|
}
|
1477 |
|
|
|
1478 |
|
|
if (new_active) {
|
1479 |
|
|
bond_set_slave_active_flags(new_active);
|
1480 |
|
|
}
|
1481 |
|
|
}
|
1482 |
|
|
|
1483 |
|
|
if (USES_PRIMARY(bond->params.mode)) {
|
1484 |
|
|
bond_mc_swap(bond, new_active, old_active);
|
1485 |
|
|
}
|
1486 |
|
|
|
1487 |
|
|
if ((bond->params.mode == BOND_MODE_TLB) ||
|
1488 |
|
|
(bond->params.mode == BOND_MODE_ALB)) {
|
1489 |
|
|
bond_alb_handle_active_change(bond, new_active);
|
1490 |
|
|
} else {
|
1491 |
|
|
bond->curr_active_slave = new_active;
|
1492 |
|
|
}
|
1493 |
|
|
}
|
1494 |
|
|
|
1495 |
|
|
/**
|
1496 |
|
|
* bond_select_active_slave - select a new active slave, if needed
|
1497 |
|
|
* @bond: our bonding struct
|
1498 |
|
|
*
|
1499 |
|
|
* This functions shoud be called when one of the following occurs:
|
1500 |
|
|
* - The old curr_active_slave has been released or lost its link.
|
1501 |
|
|
* - The primary_slave has got its link back.
|
1502 |
|
|
* - A slave has got its link back and there's no old curr_active_slave.
|
1503 |
|
|
*
|
1504 |
|
|
* Warning: Caller must hold curr_slave_lock for writing.
|
1505 |
|
|
*/
|
1506 |
|
|
static void bond_select_active_slave(struct bonding *bond)
|
1507 |
|
|
{
|
1508 |
|
|
struct slave *best_slave;
|
1509 |
|
|
|
1510 |
|
|
best_slave = bond_find_best_slave(bond);
|
1511 |
|
|
if (best_slave != bond->curr_active_slave) {
|
1512 |
|
|
bond_change_active_slave(bond, best_slave);
|
1513 |
|
|
}
|
1514 |
|
|
}
|
1515 |
|
|
|
1516 |
|
|
/*--------------------------- slave list handling ---------------------------*/
|
1517 |
|
|
|
1518 |
|
|
/*
|
1519 |
|
|
* This function attaches the slave to the end of list.
|
1520 |
|
|
*
|
1521 |
|
|
* bond->lock held for writing by caller.
|
1522 |
|
|
*/
|
1523 |
|
|
static void bond_attach_slave(struct bonding *bond, struct slave *new_slave)
|
1524 |
|
|
{
|
1525 |
|
|
if (bond->first_slave == NULL) { /* attaching the first slave */
|
1526 |
|
|
new_slave->next = new_slave;
|
1527 |
|
|
new_slave->prev = new_slave;
|
1528 |
|
|
bond->first_slave = new_slave;
|
1529 |
|
|
} else {
|
1530 |
|
|
new_slave->next = bond->first_slave;
|
1531 |
|
|
new_slave->prev = bond->first_slave->prev;
|
1532 |
|
|
new_slave->next->prev = new_slave;
|
1533 |
|
|
new_slave->prev->next = new_slave;
|
1534 |
|
|
}
|
1535 |
|
|
|
1536 |
|
|
bond->slave_cnt++;
|
1537 |
|
|
}
|
1538 |
|
|
|
1539 |
|
|
/*
|
1540 |
|
|
* This function detaches the slave from the list.
|
1541 |
|
|
* WARNING: no check is made to verify if the slave effectively
|
1542 |
|
|
* belongs to <bond>.
|
1543 |
|
|
* Nothing is freed on return, structures are just unchained.
|
1544 |
|
|
* If any slave pointer in bond was pointing to <slave>,
|
1545 |
|
|
* it should be changed by the calling function.
|
1546 |
|
|
*
|
1547 |
|
|
* bond->lock held for writing by caller.
|
1548 |
|
|
*/
|
1549 |
|
|
static void bond_detach_slave(struct bonding *bond, struct slave *slave)
|
1550 |
|
|
{
|
1551 |
|
|
if (slave->next) {
|
1552 |
|
|
slave->next->prev = slave->prev;
|
1553 |
|
|
}
|
1554 |
|
|
|
1555 |
|
|
if (slave->prev) {
|
1556 |
|
|
slave->prev->next = slave->next;
|
1557 |
|
|
}
|
1558 |
|
|
|
1559 |
|
|
if (bond->first_slave == slave) { /* slave is the first slave */
|
1560 |
|
|
if (bond->slave_cnt > 1) { /* there are more slave */
|
1561 |
|
|
bond->first_slave = slave->next;
|
1562 |
|
|
} else {
|
1563 |
|
|
bond->first_slave = NULL; /* slave was the last one */
|
1564 |
|
|
}
|
1565 |
|
|
}
|
1566 |
|
|
|
1567 |
|
|
slave->next = NULL;
|
1568 |
|
|
slave->prev = NULL;
|
1569 |
|
|
bond->slave_cnt--;
|
1570 |
|
|
}
|
1571 |
|
|
|
1572 |
|
|
/*---------------------------------- IOCTL ----------------------------------*/
|
1573 |
|
|
|
1574 |
|
|
static int bond_sethwaddr(struct net_device *bond_dev, struct net_device *slave_dev)
|
1575 |
|
|
{
|
1576 |
|
|
dprintk("bond_dev=%p\n", bond_dev);
|
1577 |
|
|
dprintk("slave_dev=%p\n", slave_dev);
|
1578 |
|
|
dprintk("slave_dev->addr_len=%d\n", slave_dev->addr_len);
|
1579 |
|
|
memcpy(bond_dev->dev_addr, slave_dev->dev_addr, slave_dev->addr_len);
|
1580 |
|
|
return 0;
|
1581 |
|
|
}
|
1582 |
|
|
|
1583 |
|
|
/* enslave device <slave> to bond device <master> */
|
1584 |
|
|
static int bond_enslave(struct net_device *bond_dev, struct net_device *slave_dev)
|
1585 |
|
|
{
|
1586 |
|
|
struct bonding *bond = bond_dev->priv;
|
1587 |
|
|
struct slave *new_slave = NULL;
|
1588 |
|
|
struct dev_mc_list *dmi;
|
1589 |
|
|
struct sockaddr addr;
|
1590 |
|
|
int link_reporting;
|
1591 |
|
|
int old_features = bond_dev->features;
|
1592 |
|
|
int res = 0;
|
1593 |
|
|
|
1594 |
|
|
if (slave_dev->do_ioctl == NULL) {
|
1595 |
|
|
printk(KERN_WARNING DRV_NAME
|
1596 |
|
|
": Warning : no link monitoring support for %s\n",
|
1597 |
|
|
slave_dev->name);
|
1598 |
|
|
}
|
1599 |
|
|
|
1600 |
|
|
/* bond must be initialized by bond_open() before enslaving */
|
1601 |
|
|
if (!(bond_dev->flags & IFF_UP)) {
|
1602 |
|
|
dprintk("Error, master_dev is not up\n");
|
1603 |
|
|
return -EPERM;
|
1604 |
|
|
}
|
1605 |
|
|
|
1606 |
|
|
/* already enslaved */
|
1607 |
|
|
if (slave_dev->flags & IFF_SLAVE) {
|
1608 |
|
|
dprintk("Error, Device was already enslaved\n");
|
1609 |
|
|
return -EBUSY;
|
1610 |
|
|
}
|
1611 |
|
|
|
1612 |
|
|
/* vlan challenged mutual exclusion */
|
1613 |
|
|
/* no need to lock since we're protected by rtnl_lock */
|
1614 |
|
|
if (slave_dev->features & NETIF_F_VLAN_CHALLENGED) {
|
1615 |
|
|
dprintk("%s: NETIF_F_VLAN_CHALLENGED\n", slave_dev->name);
|
1616 |
|
|
if (!list_empty(&bond->vlan_list)) {
|
1617 |
|
|
printk(KERN_ERR DRV_NAME
|
1618 |
|
|
": Error: cannot enslave VLAN "
|
1619 |
|
|
"challenged slave %s on VLAN enabled "
|
1620 |
|
|
"bond %s\n", slave_dev->name,
|
1621 |
|
|
bond_dev->name);
|
1622 |
|
|
return -EPERM;
|
1623 |
|
|
} else {
|
1624 |
|
|
printk(KERN_WARNING DRV_NAME
|
1625 |
|
|
": Warning: enslaved VLAN challenged "
|
1626 |
|
|
"slave %s. Adding VLANs will be blocked as "
|
1627 |
|
|
"long as %s is part of bond %s\n",
|
1628 |
|
|
slave_dev->name, slave_dev->name,
|
1629 |
|
|
bond_dev->name);
|
1630 |
|
|
bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
|
1631 |
|
|
}
|
1632 |
|
|
} else {
|
1633 |
|
|
dprintk("%s: ! NETIF_F_VLAN_CHALLENGED\n", slave_dev->name);
|
1634 |
|
|
if (bond->slave_cnt == 0) {
|
1635 |
|
|
/* First slave, and it is not VLAN challenged,
|
1636 |
|
|
* so remove the block of adding VLANs over the bond.
|
1637 |
|
|
*/
|
1638 |
|
|
bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED;
|
1639 |
|
|
}
|
1640 |
|
|
}
|
1641 |
|
|
|
1642 |
|
|
if (app_abi_ver >= 1) {
|
1643 |
|
|
/* The application is using an ABI, which requires the
|
1644 |
|
|
* slave interface to be closed.
|
1645 |
|
|
*/
|
1646 |
|
|
if ((slave_dev->flags & IFF_UP)) {
|
1647 |
|
|
printk(KERN_ERR DRV_NAME
|
1648 |
|
|
": Error: %s is up\n",
|
1649 |
|
|
slave_dev->name);
|
1650 |
|
|
res = -EPERM;
|
1651 |
|
|
goto err_undo_flags;
|
1652 |
|
|
}
|
1653 |
|
|
|
1654 |
|
|
if (slave_dev->set_mac_address == NULL) {
|
1655 |
|
|
printk(KERN_ERR DRV_NAME
|
1656 |
|
|
": Error: The slave device you specified does "
|
1657 |
|
|
"not support setting the MAC address.\n");
|
1658 |
|
|
printk(KERN_ERR
|
1659 |
|
|
"Your kernel likely does not support slave "
|
1660 |
|
|
"devices.\n");
|
1661 |
|
|
|
1662 |
|
|
res = -EOPNOTSUPP;
|
1663 |
|
|
goto err_undo_flags;
|
1664 |
|
|
}
|
1665 |
|
|
} else {
|
1666 |
|
|
/* The application is not using an ABI, which requires the
|
1667 |
|
|
* slave interface to be open.
|
1668 |
|
|
*/
|
1669 |
|
|
if (!(slave_dev->flags & IFF_UP)) {
|
1670 |
|
|
printk(KERN_ERR DRV_NAME
|
1671 |
|
|
": Error: %s is not running\n",
|
1672 |
|
|
slave_dev->name);
|
1673 |
|
|
res = -EINVAL;
|
1674 |
|
|
goto err_undo_flags;
|
1675 |
|
|
}
|
1676 |
|
|
|
1677 |
|
|
if ((bond->params.mode == BOND_MODE_8023AD) ||
|
1678 |
|
|
(bond->params.mode == BOND_MODE_TLB) ||
|
1679 |
|
|
(bond->params.mode == BOND_MODE_ALB)) {
|
1680 |
|
|
printk(KERN_ERR DRV_NAME
|
1681 |
|
|
": Error: to use %s mode, you must upgrade "
|
1682 |
|
|
"ifenslave.\n",
|
1683 |
|
|
bond_mode_name(bond->params.mode));
|
1684 |
|
|
res = -EOPNOTSUPP;
|
1685 |
|
|
goto err_undo_flags;
|
1686 |
|
|
}
|
1687 |
|
|
}
|
1688 |
|
|
|
1689 |
|
|
new_slave = kmalloc(sizeof(struct slave), GFP_KERNEL);
|
1690 |
|
|
if (!new_slave) {
|
1691 |
|
|
res = -ENOMEM;
|
1692 |
|
|
goto err_undo_flags;
|
1693 |
|
|
}
|
1694 |
|
|
|
1695 |
|
|
memset(new_slave, 0, sizeof(struct slave));
|
1696 |
|
|
|
1697 |
|
|
/* save slave's original flags before calling
|
1698 |
|
|
* netdev_set_master and dev_open
|
1699 |
|
|
*/
|
1700 |
|
|
new_slave->original_flags = slave_dev->flags;
|
1701 |
|
|
|
1702 |
|
|
if (app_abi_ver >= 1) {
|
1703 |
|
|
/* save slave's original ("permanent") mac address for
|
1704 |
|
|
* modes that needs it, and for restoring it upon release,
|
1705 |
|
|
* and then set it to the master's address
|
1706 |
|
|
*/
|
1707 |
|
|
memcpy(new_slave->perm_hwaddr, slave_dev->dev_addr, ETH_ALEN);
|
1708 |
|
|
|
1709 |
|
|
/* set slave to master's mac address
|
1710 |
|
|
* The application already set the master's
|
1711 |
|
|
* mac address to that of the first slave
|
1712 |
|
|
*/
|
1713 |
|
|
memcpy(addr.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
|
1714 |
|
|
addr.sa_family = slave_dev->type;
|
1715 |
|
|
res = slave_dev->set_mac_address(slave_dev, &addr);
|
1716 |
|
|
if (res) {
|
1717 |
|
|
dprintk("Error %d calling set_mac_address\n", res);
|
1718 |
|
|
goto err_free;
|
1719 |
|
|
}
|
1720 |
|
|
|
1721 |
|
|
/* open the slave since the application closed it */
|
1722 |
|
|
res = dev_open(slave_dev);
|
1723 |
|
|
if (res) {
|
1724 |
|
|
dprintk("Openning slave %s failed\n", slave_dev->name);
|
1725 |
|
|
goto err_restore_mac;
|
1726 |
|
|
}
|
1727 |
|
|
}
|
1728 |
|
|
|
1729 |
|
|
res = netdev_set_master(slave_dev, bond_dev);
|
1730 |
|
|
if (res) {
|
1731 |
|
|
dprintk("Error %d calling netdev_set_master\n", res);
|
1732 |
|
|
if (app_abi_ver < 1) {
|
1733 |
|
|
goto err_free;
|
1734 |
|
|
} else {
|
1735 |
|
|
goto err_close;
|
1736 |
|
|
}
|
1737 |
|
|
}
|
1738 |
|
|
|
1739 |
|
|
new_slave->dev = slave_dev;
|
1740 |
|
|
|
1741 |
|
|
if ((bond->params.mode == BOND_MODE_TLB) ||
|
1742 |
|
|
(bond->params.mode == BOND_MODE_ALB)) {
|
1743 |
|
|
/* bond_alb_init_slave() must be called before all other stages since
|
1744 |
|
|
* it might fail and we do not want to have to undo everything
|
1745 |
|
|
*/
|
1746 |
|
|
res = bond_alb_init_slave(bond, new_slave);
|
1747 |
|
|
if (res) {
|
1748 |
|
|
goto err_unset_master;
|
1749 |
|
|
}
|
1750 |
|
|
}
|
1751 |
|
|
|
1752 |
|
|
/* If the mode USES_PRIMARY, then the new slave gets the
|
1753 |
|
|
* master's promisc (and mc) settings only if it becomes the
|
1754 |
|
|
* curr_active_slave, and that is taken care of later when calling
|
1755 |
|
|
* bond_change_active()
|
1756 |
|
|
*/
|
1757 |
|
|
if (!USES_PRIMARY(bond->params.mode)) {
|
1758 |
|
|
/* set promiscuity level to new slave */
|
1759 |
|
|
if (bond_dev->flags & IFF_PROMISC) {
|
1760 |
|
|
dev_set_promiscuity(slave_dev, 1);
|
1761 |
|
|
}
|
1762 |
|
|
|
1763 |
|
|
/* set allmulti level to new slave */
|
1764 |
|
|
if (bond_dev->flags & IFF_ALLMULTI) {
|
1765 |
|
|
dev_set_allmulti(slave_dev, 1);
|
1766 |
|
|
}
|
1767 |
|
|
|
1768 |
|
|
/* upload master's mc_list to new slave */
|
1769 |
|
|
for (dmi = bond_dev->mc_list; dmi; dmi = dmi->next) {
|
1770 |
|
|
dev_mc_add (slave_dev, dmi->dmi_addr, dmi->dmi_addrlen, 0);
|
1771 |
|
|
}
|
1772 |
|
|
}
|
1773 |
|
|
|
1774 |
|
|
if (bond->params.mode == BOND_MODE_8023AD) {
|
1775 |
|
|
/* add lacpdu mc addr to mc list */
|
1776 |
|
|
u8 lacpdu_multicast[ETH_ALEN] = MULTICAST_LACPDU_ADDR;
|
1777 |
|
|
|
1778 |
|
|
dev_mc_add(slave_dev, lacpdu_multicast, ETH_ALEN, 0);
|
1779 |
|
|
}
|
1780 |
|
|
|
1781 |
|
|
bond_add_vlans_on_slave(bond, slave_dev);
|
1782 |
|
|
|
1783 |
|
|
write_lock_bh(&bond->lock);
|
1784 |
|
|
|
1785 |
|
|
bond_attach_slave(bond, new_slave);
|
1786 |
|
|
|
1787 |
|
|
new_slave->delay = 0;
|
1788 |
|
|
new_slave->link_failure_count = 0;
|
1789 |
|
|
|
1790 |
|
|
if (bond->params.miimon && !bond->params.use_carrier) {
|
1791 |
|
|
link_reporting = bond_check_dev_link(bond, slave_dev, 1);
|
1792 |
|
|
|
1793 |
|
|
if ((link_reporting == -1) && !bond->params.arp_interval) {
|
1794 |
|
|
/*
|
1795 |
|
|
* miimon is set but a bonded network driver
|
1796 |
|
|
* does not support ETHTOOL/MII and
|
1797 |
|
|
* arp_interval is not set. Note: if
|
1798 |
|
|
* use_carrier is enabled, we will never go
|
1799 |
|
|
* here (because netif_carrier is always
|
1800 |
|
|
* supported); thus, we don't need to change
|
1801 |
|
|
* the messages for netif_carrier.
|
1802 |
|
|
*/
|
1803 |
|
|
printk(KERN_WARNING DRV_NAME
|
1804 |
|
|
": Warning: MII and ETHTOOL support not "
|
1805 |
|
|
"available for interface %s, and "
|
1806 |
|
|
"arp_interval/arp_ip_target module parameters "
|
1807 |
|
|
"not specified, thus bonding will not detect "
|
1808 |
|
|
"link failures! see bonding.txt for details.\n",
|
1809 |
|
|
slave_dev->name);
|
1810 |
|
|
} else if (link_reporting == -1) {
|
1811 |
|
|
/* unable get link status using mii/ethtool */
|
1812 |
|
|
printk(KERN_WARNING DRV_NAME
|
1813 |
|
|
": Warning: can't get link status from "
|
1814 |
|
|
"interface %s; the network driver associated "
|
1815 |
|
|
"with this interface does not support MII or "
|
1816 |
|
|
"ETHTOOL link status reporting, thus miimon "
|
1817 |
|
|
"has no effect on this interface.\n",
|
1818 |
|
|
slave_dev->name);
|
1819 |
|
|
}
|
1820 |
|
|
}
|
1821 |
|
|
|
1822 |
|
|
/* check for initial state */
|
1823 |
|
|
if (!bond->params.miimon ||
|
1824 |
|
|
(bond_check_dev_link(bond, slave_dev, 0) == BMSR_LSTATUS)) {
|
1825 |
|
|
if (bond->params.updelay) {
|
1826 |
|
|
dprintk("Initial state of slave_dev is "
|
1827 |
|
|
"BOND_LINK_BACK\n");
|
1828 |
|
|
new_slave->link = BOND_LINK_BACK;
|
1829 |
|
|
new_slave->delay = bond->params.updelay;
|
1830 |
|
|
} else {
|
1831 |
|
|
dprintk("Initial state of slave_dev is "
|
1832 |
|
|
"BOND_LINK_UP\n");
|
1833 |
|
|
new_slave->link = BOND_LINK_UP;
|
1834 |
|
|
}
|
1835 |
|
|
new_slave->jiffies = jiffies;
|
1836 |
|
|
} else {
|
1837 |
|
|
dprintk("Initial state of slave_dev is "
|
1838 |
|
|
"BOND_LINK_DOWN\n");
|
1839 |
|
|
new_slave->link = BOND_LINK_DOWN;
|
1840 |
|
|
}
|
1841 |
|
|
|
1842 |
|
|
if (bond_update_speed_duplex(new_slave) &&
|
1843 |
|
|
(new_slave->link != BOND_LINK_DOWN)) {
|
1844 |
|
|
printk(KERN_WARNING DRV_NAME
|
1845 |
|
|
": Warning: failed to get speed/duplex from %s, speed "
|
1846 |
|
|
"forced to 100Mbps, duplex forced to Full.\n",
|
1847 |
|
|
new_slave->dev->name);
|
1848 |
|
|
|
1849 |
|
|
if (bond->params.mode == BOND_MODE_8023AD) {
|
1850 |
|
|
printk(KERN_WARNING
|
1851 |
|
|
"Operation of 802.3ad mode requires ETHTOOL "
|
1852 |
|
|
"support in base driver for proper aggregator "
|
1853 |
|
|
"selection.\n");
|
1854 |
|
|
}
|
1855 |
|
|
}
|
1856 |
|
|
|
1857 |
|
|
if (USES_PRIMARY(bond->params.mode) && bond->params.primary[0]) {
|
1858 |
|
|
/* if there is a primary slave, remember it */
|
1859 |
|
|
if (strcmp(bond->params.primary, new_slave->dev->name) == 0) {
|
1860 |
|
|
bond->primary_slave = new_slave;
|
1861 |
|
|
}
|
1862 |
|
|
}
|
1863 |
|
|
|
1864 |
|
|
switch (bond->params.mode) {
|
1865 |
|
|
case BOND_MODE_ACTIVEBACKUP:
|
1866 |
|
|
/* if we're in active-backup mode, we need one and only one active
|
1867 |
|
|
* interface. The backup interfaces will have their NOARP flag set
|
1868 |
|
|
* because we need them to be completely deaf and not to respond to
|
1869 |
|
|
* any ARP request on the network to avoid fooling a switch. Thus,
|
1870 |
|
|
* since we guarantee that curr_active_slave always point to the last
|
1871 |
|
|
* usable interface, we just have to verify this interface's flag.
|
1872 |
|
|
*/
|
1873 |
|
|
if (((!bond->curr_active_slave) ||
|
1874 |
|
|
(bond->curr_active_slave->dev->flags & IFF_NOARP)) &&
|
1875 |
|
|
(new_slave->link != BOND_LINK_DOWN)) {
|
1876 |
|
|
dprintk("This is the first active slave\n");
|
1877 |
|
|
/* first slave or no active slave yet, and this link
|
1878 |
|
|
is OK, so make this interface the active one */
|
1879 |
|
|
bond_change_active_slave(bond, new_slave);
|
1880 |
|
|
} else {
|
1881 |
|
|
dprintk("This is just a backup slave\n");
|
1882 |
|
|
bond_set_slave_inactive_flags(new_slave);
|
1883 |
|
|
}
|
1884 |
|
|
break;
|
1885 |
|
|
case BOND_MODE_8023AD:
|
1886 |
|
|
/* in 802.3ad mode, the internal mechanism
|
1887 |
|
|
* will activate the slaves in the selected
|
1888 |
|
|
* aggregator
|
1889 |
|
|
*/
|
1890 |
|
|
bond_set_slave_inactive_flags(new_slave);
|
1891 |
|
|
/* if this is the first slave */
|
1892 |
|
|
if (bond->slave_cnt == 1) {
|
1893 |
|
|
SLAVE_AD_INFO(new_slave).id = 1;
|
1894 |
|
|
/* Initialize AD with the number of times that the AD timer is called in 1 second
|
1895 |
|
|
* can be called only after the mac address of the bond is set
|
1896 |
|
|
*/
|
1897 |
|
|
bond_3ad_initialize(bond, 1000/AD_TIMER_INTERVAL,
|
1898 |
|
|
bond->params.lacp_fast);
|
1899 |
|
|
} else {
|
1900 |
|
|
SLAVE_AD_INFO(new_slave).id =
|
1901 |
|
|
SLAVE_AD_INFO(new_slave->prev).id + 1;
|
1902 |
|
|
}
|
1903 |
|
|
|
1904 |
|
|
bond_3ad_bind_slave(new_slave);
|
1905 |
|
|
break;
|
1906 |
|
|
case BOND_MODE_TLB:
|
1907 |
|
|
case BOND_MODE_ALB:
|
1908 |
|
|
new_slave->state = BOND_STATE_ACTIVE;
|
1909 |
|
|
if ((!bond->curr_active_slave) &&
|
1910 |
|
|
(new_slave->link != BOND_LINK_DOWN)) {
|
1911 |
|
|
/* first slave or no active slave yet, and this link
|
1912 |
|
|
* is OK, so make this interface the active one
|
1913 |
|
|
*/
|
1914 |
|
|
bond_change_active_slave(bond, new_slave);
|
1915 |
|
|
}
|
1916 |
|
|
break;
|
1917 |
|
|
default:
|
1918 |
|
|
dprintk("This slave is always active in trunk mode\n");
|
1919 |
|
|
|
1920 |
|
|
/* always active in trunk mode */
|
1921 |
|
|
new_slave->state = BOND_STATE_ACTIVE;
|
1922 |
|
|
|
1923 |
|
|
/* In trunking mode there is little meaning to curr_active_slave
|
1924 |
|
|
* anyway (it holds no special properties of the bond device),
|
1925 |
|
|
* so we can change it without calling change_active_interface()
|
1926 |
|
|
*/
|
1927 |
|
|
if (!bond->curr_active_slave) {
|
1928 |
|
|
bond->curr_active_slave = new_slave;
|
1929 |
|
|
}
|
1930 |
|
|
break;
|
1931 |
|
|
} /* switch(bond_mode) */
|
1932 |
|
|
|
1933 |
|
|
write_unlock_bh(&bond->lock);
|
1934 |
|
|
|
1935 |
|
|
if (app_abi_ver < 1) {
|
1936 |
|
|
/*
|
1937 |
|
|
* !!! This is to support old versions of ifenslave.
|
1938 |
|
|
* We can remove this in 2.5 because our ifenslave takes
|
1939 |
|
|
* care of this for us.
|
1940 |
|
|
* We check to see if the master has a mac address yet.
|
1941 |
|
|
* If not, we'll give it the mac address of our slave device.
|
1942 |
|
|
*/
|
1943 |
|
|
int ndx = 0;
|
1944 |
|
|
|
1945 |
|
|
for (ndx = 0; ndx < bond_dev->addr_len; ndx++) {
|
1946 |
|
|
dprintk("Checking ndx=%d of bond_dev->dev_addr\n",
|
1947 |
|
|
ndx);
|
1948 |
|
|
if (bond_dev->dev_addr[ndx] != 0) {
|
1949 |
|
|
dprintk("Found non-zero byte at ndx=%d\n",
|
1950 |
|
|
ndx);
|
1951 |
|
|
break;
|
1952 |
|
|
}
|
1953 |
|
|
}
|
1954 |
|
|
|
1955 |
|
|
if (ndx == bond_dev->addr_len) {
|
1956 |
|
|
/*
|
1957 |
|
|
* We got all the way through the address and it was
|
1958 |
|
|
* all 0's.
|
1959 |
|
|
*/
|
1960 |
|
|
dprintk("%s doesn't have a MAC address yet. \n",
|
1961 |
|
|
bond_dev->name);
|
1962 |
|
|
dprintk("Going to give assign it from %s.\n",
|
1963 |
|
|
slave_dev->name);
|
1964 |
|
|
bond_sethwaddr(bond_dev, slave_dev);
|
1965 |
|
|
}
|
1966 |
|
|
}
|
1967 |
|
|
|
1968 |
|
|
printk(KERN_INFO DRV_NAME
|
1969 |
|
|
": %s: enslaving %s as a%s interface with a%s link.\n",
|
1970 |
|
|
bond_dev->name, slave_dev->name,
|
1971 |
|
|
new_slave->state == BOND_STATE_ACTIVE ? "n active" : " backup",
|
1972 |
|
|
new_slave->link != BOND_LINK_DOWN ? "n up" : " down");
|
1973 |
|
|
|
1974 |
|
|
/* enslave is successful */
|
1975 |
|
|
return 0;
|
1976 |
|
|
|
1977 |
|
|
/* Undo stages on error */
|
1978 |
|
|
err_unset_master:
|
1979 |
|
|
netdev_set_master(slave_dev, NULL);
|
1980 |
|
|
|
1981 |
|
|
err_close:
|
1982 |
|
|
dev_close(slave_dev);
|
1983 |
|
|
|
1984 |
|
|
err_restore_mac:
|
1985 |
|
|
memcpy(addr.sa_data, new_slave->perm_hwaddr, ETH_ALEN);
|
1986 |
|
|
addr.sa_family = slave_dev->type;
|
1987 |
|
|
slave_dev->set_mac_address(slave_dev, &addr);
|
1988 |
|
|
|
1989 |
|
|
err_free:
|
1990 |
|
|
kfree(new_slave);
|
1991 |
|
|
|
1992 |
|
|
err_undo_flags:
|
1993 |
|
|
bond_dev->features = old_features;
|
1994 |
|
|
|
1995 |
|
|
return res;
|
1996 |
|
|
}
|
1997 |
|
|
|
1998 |
|
|
/*
|
1999 |
|
|
* Try to release the slave device <slave> from the bond device <master>
|
2000 |
|
|
* It is legal to access curr_active_slave without a lock because all the function
|
2001 |
|
|
* is write-locked.
|
2002 |
|
|
*
|
2003 |
|
|
* The rules for slave state should be:
|
2004 |
|
|
* for Active/Backup:
|
2005 |
|
|
* Active stays on all backups go down
|
2006 |
|
|
* for Bonded connections:
|
2007 |
|
|
* The first up interface should be left on and all others downed.
|
2008 |
|
|
*/
|
2009 |
|
|
static int bond_release(struct net_device *bond_dev, struct net_device *slave_dev)
|
2010 |
|
|
{
|
2011 |
|
|
struct bonding *bond = bond_dev->priv;
|
2012 |
|
|
struct slave *slave, *oldcurrent;
|
2013 |
|
|
struct sockaddr addr;
|
2014 |
|
|
int mac_addr_differ;
|
2015 |
|
|
|
2016 |
|
|
/* slave is not a slave or master is not master of this slave */
|
2017 |
|
|
if (!(slave_dev->flags & IFF_SLAVE) ||
|
2018 |
|
|
(slave_dev->master != bond_dev)) {
|
2019 |
|
|
printk(KERN_ERR DRV_NAME
|
2020 |
|
|
": Error: %s: cannot release %s.\n",
|
2021 |
|
|
bond_dev->name, slave_dev->name);
|
2022 |
|
|
return -EINVAL;
|
2023 |
|
|
}
|
2024 |
|
|
|
2025 |
|
|
write_lock_bh(&bond->lock);
|
2026 |
|
|
|
2027 |
|
|
slave = bond_get_slave_by_dev(bond, slave_dev);
|
2028 |
|
|
if (!slave) {
|
2029 |
|
|
/* not a slave of this bond */
|
2030 |
|
|
printk(KERN_INFO DRV_NAME
|
2031 |
|
|
": %s: %s not enslaved\n",
|
2032 |
|
|
bond_dev->name, slave_dev->name);
|
2033 |
|
|
return -EINVAL;
|
2034 |
|
|
}
|
2035 |
|
|
|
2036 |
|
|
mac_addr_differ = memcmp(bond_dev->dev_addr,
|
2037 |
|
|
slave->perm_hwaddr,
|
2038 |
|
|
ETH_ALEN);
|
2039 |
|
|
if (!mac_addr_differ && (bond->slave_cnt > 1)) {
|
2040 |
|
|
printk(KERN_WARNING DRV_NAME
|
2041 |
|
|
": Warning: the permanent HWaddr of %s "
|
2042 |
|
|
"- %02X:%02X:%02X:%02X:%02X:%02X - is "
|
2043 |
|
|
"still in use by %s. Set the HWaddr of "
|
2044 |
|
|
"%s to a different address to avoid "
|
2045 |
|
|
"conflicts.\n",
|
2046 |
|
|
slave_dev->name,
|
2047 |
|
|
slave->perm_hwaddr[0],
|
2048 |
|
|
slave->perm_hwaddr[1],
|
2049 |
|
|
slave->perm_hwaddr[2],
|
2050 |
|
|
slave->perm_hwaddr[3],
|
2051 |
|
|
slave->perm_hwaddr[4],
|
2052 |
|
|
slave->perm_hwaddr[5],
|
2053 |
|
|
bond_dev->name,
|
2054 |
|
|
slave_dev->name);
|
2055 |
|
|
}
|
2056 |
|
|
|
2057 |
|
|
/* Inform AD package of unbinding of slave. */
|
2058 |
|
|
if (bond->params.mode == BOND_MODE_8023AD) {
|
2059 |
|
|
/* must be called before the slave is
|
2060 |
|
|
* detached from the list
|
2061 |
|
|
*/
|
2062 |
|
|
bond_3ad_unbind_slave(slave);
|
2063 |
|
|
}
|
2064 |
|
|
|
2065 |
|
|
printk(KERN_INFO DRV_NAME
|
2066 |
|
|
": %s: releasing %s interface %s\n",
|
2067 |
|
|
bond_dev->name,
|
2068 |
|
|
(slave->state == BOND_STATE_ACTIVE)
|
2069 |
|
|
? "active" : "backup",
|
2070 |
|
|
slave_dev->name);
|
2071 |
|
|
|
2072 |
|
|
oldcurrent = bond->curr_active_slave;
|
2073 |
|
|
|
2074 |
|
|
bond->current_arp_slave = NULL;
|
2075 |
|
|
|
2076 |
|
|
/* release the slave from its bond */
|
2077 |
|
|
bond_detach_slave(bond, slave);
|
2078 |
|
|
|
2079 |
|
|
if (bond->primary_slave == slave) {
|
2080 |
|
|
bond->primary_slave = NULL;
|
2081 |
|
|
}
|
2082 |
|
|
|
2083 |
|
|
if (oldcurrent == slave) {
|
2084 |
|
|
bond_change_active_slave(bond, NULL);
|
2085 |
|
|
}
|
2086 |
|
|
|
2087 |
|
|
if ((bond->params.mode == BOND_MODE_TLB) ||
|
2088 |
|
|
(bond->params.mode == BOND_MODE_ALB)) {
|
2089 |
|
|
/* Must be called only after the slave has been
|
2090 |
|
|
* detached from the list and the curr_active_slave
|
2091 |
|
|
* has been cleared (if our_slave == old_current),
|
2092 |
|
|
* but before a new active slave is selected.
|
2093 |
|
|
*/
|
2094 |
|
|
bond_alb_deinit_slave(bond, slave);
|
2095 |
|
|
}
|
2096 |
|
|
|
2097 |
|
|
if (oldcurrent == slave) {
|
2098 |
|
|
bond_select_active_slave(bond);
|
2099 |
|
|
|
2100 |
|
|
if (!bond->curr_active_slave) {
|
2101 |
|
|
printk(KERN_INFO DRV_NAME
|
2102 |
|
|
": %s: now running without any active "
|
2103 |
|
|
"interface !\n",
|
2104 |
|
|
bond_dev->name);
|
2105 |
|
|
}
|
2106 |
|
|
}
|
2107 |
|
|
|
2108 |
|
|
if (bond->slave_cnt == 0) {
|
2109 |
|
|
/* if the last slave was removed, zero the mac address
|
2110 |
|
|
* of the master so it will be set by the application
|
2111 |
|
|
* to the mac address of the first slave
|
2112 |
|
|
*/
|
2113 |
|
|
memset(bond_dev->dev_addr, 0, bond_dev->addr_len);
|
2114 |
|
|
|
2115 |
|
|
if (list_empty(&bond->vlan_list)) {
|
2116 |
|
|
bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
|
2117 |
|
|
} else {
|
2118 |
|
|
printk(KERN_WARNING DRV_NAME
|
2119 |
|
|
": Warning: clearing HW address of %s while it "
|
2120 |
|
|
"still has VLANs.\n",
|
2121 |
|
|
bond_dev->name);
|
2122 |
|
|
printk(KERN_WARNING DRV_NAME
|
2123 |
|
|
": When re-adding slaves, make sure the bond's "
|
2124 |
|
|
"HW address matches its VLANs'.\n");
|
2125 |
|
|
}
|
2126 |
|
|
} else if ((bond_dev->features & NETIF_F_VLAN_CHALLENGED) &&
|
2127 |
|
|
!bond_has_challenged_slaves(bond)) {
|
2128 |
|
|
printk(KERN_INFO DRV_NAME
|
2129 |
|
|
": last VLAN challenged slave %s "
|
2130 |
|
|
"left bond %s. VLAN blocking is removed\n",
|
2131 |
|
|
slave_dev->name, bond_dev->name);
|
2132 |
|
|
bond_dev->features &= ~NETIF_F_VLAN_CHALLENGED;
|
2133 |
|
|
}
|
2134 |
|
|
|
2135 |
|
|
write_unlock_bh(&bond->lock);
|
2136 |
|
|
|
2137 |
|
|
bond_del_vlans_from_slave(bond, slave_dev);
|
2138 |
|
|
|
2139 |
|
|
/* If the mode USES_PRIMARY, then we should only remove its
|
2140 |
|
|
* promisc and mc settings if it was the curr_active_slave, but that was
|
2141 |
|
|
* already taken care of above when we detached the slave
|
2142 |
|
|
*/
|
2143 |
|
|
if (!USES_PRIMARY(bond->params.mode)) {
|
2144 |
|
|
/* unset promiscuity level from slave */
|
2145 |
|
|
if (bond_dev->flags & IFF_PROMISC) {
|
2146 |
|
|
dev_set_promiscuity(slave_dev, -1);
|
2147 |
|
|
}
|
2148 |
|
|
|
2149 |
|
|
/* unset allmulti level from slave */
|
2150 |
|
|
if (bond_dev->flags & IFF_ALLMULTI) {
|
2151 |
|
|
dev_set_allmulti(slave_dev, -1);
|
2152 |
|
|
}
|
2153 |
|
|
|
2154 |
|
|
/* flush master's mc_list from slave */
|
2155 |
|
|
bond_mc_list_flush(bond_dev, slave_dev);
|
2156 |
|
|
}
|
2157 |
|
|
|
2158 |
|
|
netdev_set_master(slave_dev, NULL);
|
2159 |
|
|
|
2160 |
|
|
/* close slave before restoring its mac address */
|
2161 |
|
|
dev_close(slave_dev);
|
2162 |
|
|
|
2163 |
|
|
if (app_abi_ver >= 1) {
|
2164 |
|
|
/* restore original ("permanent") mac address */
|
2165 |
|
|
memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN);
|
2166 |
|
|
addr.sa_family = slave_dev->type;
|
2167 |
|
|
slave_dev->set_mac_address(slave_dev, &addr);
|
2168 |
|
|
}
|
2169 |
|
|
|
2170 |
|
|
/* restore the original state of the
|
2171 |
|
|
* IFF_NOARP flag that might have been
|
2172 |
|
|
* set by bond_set_slave_inactive_flags()
|
2173 |
|
|
*/
|
2174 |
|
|
if ((slave->original_flags & IFF_NOARP) == 0) {
|
2175 |
|
|
slave_dev->flags &= ~IFF_NOARP;
|
2176 |
|
|
}
|
2177 |
|
|
|
2178 |
|
|
kfree(slave);
|
2179 |
|
|
|
2180 |
|
|
return 0; /* deletion OK */
|
2181 |
|
|
}
|
2182 |
|
|
|
2183 |
|
|
/*
|
2184 |
|
|
* This function releases all slaves.
|
2185 |
|
|
*/
|
2186 |
|
|
static int bond_release_all(struct net_device *bond_dev)
|
2187 |
|
|
{
|
2188 |
|
|
struct bonding *bond = bond_dev->priv;
|
2189 |
|
|
struct slave *slave;
|
2190 |
|
|
struct net_device *slave_dev;
|
2191 |
|
|
struct sockaddr addr;
|
2192 |
|
|
|
2193 |
|
|
write_lock_bh(&bond->lock);
|
2194 |
|
|
|
2195 |
|
|
if (bond->slave_cnt == 0) {
|
2196 |
|
|
goto out;
|
2197 |
|
|
}
|
2198 |
|
|
|
2199 |
|
|
bond->current_arp_slave = NULL;
|
2200 |
|
|
bond->primary_slave = NULL;
|
2201 |
|
|
bond_change_active_slave(bond, NULL);
|
2202 |
|
|
|
2203 |
|
|
while ((slave = bond->first_slave) != NULL) {
|
2204 |
|
|
/* Inform AD package of unbinding of slave
|
2205 |
|
|
* before slave is detached from the list.
|
2206 |
|
|
*/
|
2207 |
|
|
if (bond->params.mode == BOND_MODE_8023AD) {
|
2208 |
|
|
bond_3ad_unbind_slave(slave);
|
2209 |
|
|
}
|
2210 |
|
|
|
2211 |
|
|
slave_dev = slave->dev;
|
2212 |
|
|
bond_detach_slave(bond, slave);
|
2213 |
|
|
|
2214 |
|
|
if ((bond->params.mode == BOND_MODE_TLB) ||
|
2215 |
|
|
(bond->params.mode == BOND_MODE_ALB)) {
|
2216 |
|
|
/* must be called only after the slave
|
2217 |
|
|
* has been detached from the list
|
2218 |
|
|
*/
|
2219 |
|
|
bond_alb_deinit_slave(bond, slave);
|
2220 |
|
|
}
|
2221 |
|
|
|
2222 |
|
|
/* now that the slave is detached, unlock and perform
|
2223 |
|
|
* all the undo steps that should not be called from
|
2224 |
|
|
* within a lock.
|
2225 |
|
|
*/
|
2226 |
|
|
write_unlock_bh(&bond->lock);
|
2227 |
|
|
|
2228 |
|
|
bond_del_vlans_from_slave(bond, slave_dev);
|
2229 |
|
|
|
2230 |
|
|
/* If the mode USES_PRIMARY, then we should only remove its
|
2231 |
|
|
* promisc and mc settings if it was the curr_active_slave, but that was
|
2232 |
|
|
* already taken care of above when we detached the slave
|
2233 |
|
|
*/
|
2234 |
|
|
if (!USES_PRIMARY(bond->params.mode)) {
|
2235 |
|
|
/* unset promiscuity level from slave */
|
2236 |
|
|
if (bond_dev->flags & IFF_PROMISC) {
|
2237 |
|
|
dev_set_promiscuity(slave_dev, -1);
|
2238 |
|
|
}
|
2239 |
|
|
|
2240 |
|
|
/* unset allmulti level from slave */
|
2241 |
|
|
if (bond_dev->flags & IFF_ALLMULTI) {
|
2242 |
|
|
dev_set_allmulti(slave_dev, -1);
|
2243 |
|
|
}
|
2244 |
|
|
|
2245 |
|
|
/* flush master's mc_list from slave */
|
2246 |
|
|
bond_mc_list_flush(bond_dev, slave_dev);
|
2247 |
|
|
}
|
2248 |
|
|
|
2249 |
|
|
netdev_set_master(slave_dev, NULL);
|
2250 |
|
|
|
2251 |
|
|
/* close slave before restoring its mac address */
|
2252 |
|
|
dev_close(slave_dev);
|
2253 |
|
|
|
2254 |
|
|
if (app_abi_ver >= 1) {
|
2255 |
|
|
/* restore original ("permanent") mac address*/
|
2256 |
|
|
memcpy(addr.sa_data, slave->perm_hwaddr, ETH_ALEN);
|
2257 |
|
|
addr.sa_family = slave_dev->type;
|
2258 |
|
|
slave_dev->set_mac_address(slave_dev, &addr);
|
2259 |
|
|
}
|
2260 |
|
|
|
2261 |
|
|
/* restore the original state of the IFF_NOARP flag that might have
|
2262 |
|
|
* been set by bond_set_slave_inactive_flags()
|
2263 |
|
|
*/
|
2264 |
|
|
if ((slave->original_flags & IFF_NOARP) == 0) {
|
2265 |
|
|
slave_dev->flags &= ~IFF_NOARP;
|
2266 |
|
|
}
|
2267 |
|
|
|
2268 |
|
|
kfree(slave);
|
2269 |
|
|
|
2270 |
|
|
/* re-acquire the lock before getting the next slave */
|
2271 |
|
|
write_lock_bh(&bond->lock);
|
2272 |
|
|
}
|
2273 |
|
|
|
2274 |
|
|
/* zero the mac address of the master so it will be
|
2275 |
|
|
* set by the application to the mac address of the
|
2276 |
|
|
* first slave
|
2277 |
|
|
*/
|
2278 |
|
|
memset(bond_dev->dev_addr, 0, bond_dev->addr_len);
|
2279 |
|
|
|
2280 |
|
|
if (list_empty(&bond->vlan_list)) {
|
2281 |
|
|
bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
|
2282 |
|
|
} else {
|
2283 |
|
|
printk(KERN_WARNING DRV_NAME
|
2284 |
|
|
": Warning: clearing HW address of %s while it "
|
2285 |
|
|
"still has VLANs.\n",
|
2286 |
|
|
bond_dev->name);
|
2287 |
|
|
printk(KERN_WARNING DRV_NAME
|
2288 |
|
|
": When re-adding slaves, make sure the bond's "
|
2289 |
|
|
"HW address matches its VLANs'.\n");
|
2290 |
|
|
}
|
2291 |
|
|
|
2292 |
|
|
printk(KERN_INFO DRV_NAME
|
2293 |
|
|
": %s: released all slaves\n",
|
2294 |
|
|
bond_dev->name);
|
2295 |
|
|
|
2296 |
|
|
out:
|
2297 |
|
|
write_unlock_bh(&bond->lock);
|
2298 |
|
|
|
2299 |
|
|
return 0;
|
2300 |
|
|
}
|
2301 |
|
|
|
2302 |
|
|
/*
|
2303 |
|
|
* This function changes the active slave to slave <slave_dev>.
|
2304 |
|
|
* It returns -EINVAL in the following cases.
|
2305 |
|
|
* - <slave_dev> is not found in the list.
|
2306 |
|
|
* - There is not active slave now.
|
2307 |
|
|
* - <slave_dev> is already active.
|
2308 |
|
|
* - The link state of <slave_dev> is not BOND_LINK_UP.
|
2309 |
|
|
* - <slave_dev> is not running.
|
2310 |
|
|
* In these cases, this fuction does nothing.
|
2311 |
|
|
* In the other cases, currnt_slave pointer is changed and 0 is returned.
|
2312 |
|
|
*/
|
2313 |
|
|
static int bond_ioctl_change_active(struct net_device *bond_dev, struct net_device *slave_dev)
|
2314 |
|
|
{
|
2315 |
|
|
struct bonding *bond = bond_dev->priv;
|
2316 |
|
|
struct slave *old_active = NULL;
|
2317 |
|
|
struct slave *new_active = NULL;
|
2318 |
|
|
int res = 0;
|
2319 |
|
|
|
2320 |
|
|
if (!USES_PRIMARY(bond->params.mode)) {
|
2321 |
|
|
return -EINVAL;
|
2322 |
|
|
}
|
2323 |
|
|
|
2324 |
|
|
/* Verify that master_dev is indeed the master of slave_dev */
|
2325 |
|
|
if (!(slave_dev->flags & IFF_SLAVE) ||
|
2326 |
|
|
(slave_dev->master != bond_dev)) {
|
2327 |
|
|
return -EINVAL;
|
2328 |
|
|
}
|
2329 |
|
|
|
2330 |
|
|
write_lock_bh(&bond->lock);
|
2331 |
|
|
|
2332 |
|
|
old_active = bond->curr_active_slave;
|
2333 |
|
|
new_active = bond_get_slave_by_dev(bond, slave_dev);
|
2334 |
|
|
|
2335 |
|
|
/*
|
2336 |
|
|
* Changing to the current active: do nothing; return success.
|
2337 |
|
|
*/
|
2338 |
|
|
if (new_active && (new_active == old_active)) {
|
2339 |
|
|
write_unlock_bh(&bond->lock);
|
2340 |
|
|
return 0;
|
2341 |
|
|
}
|
2342 |
|
|
|
2343 |
|
|
if ((new_active) &&
|
2344 |
|
|
(old_active) &&
|
2345 |
|
|
(new_active->link == BOND_LINK_UP) &&
|
2346 |
|
|
IS_UP(new_active->dev)) {
|
2347 |
|
|
bond_change_active_slave(bond, new_active);
|
2348 |
|
|
} else {
|
2349 |
|
|
res = -EINVAL;
|
2350 |
|
|
}
|
2351 |
|
|
|
2352 |
|
|
write_unlock_bh(&bond->lock);
|
2353 |
|
|
|
2354 |
|
|
return res;
|
2355 |
|
|
}
|
2356 |
|
|
|
2357 |
|
|
static int bond_ethtool_ioctl(struct net_device *bond_dev, struct ifreq *ifr)
|
2358 |
|
|
{
|
2359 |
|
|
struct ethtool_drvinfo info;
|
2360 |
|
|
void *addr = ifr->ifr_data;
|
2361 |
|
|
uint32_t cmd;
|
2362 |
|
|
|
2363 |
|
|
if (get_user(cmd, (uint32_t *)addr)) {
|
2364 |
|
|
return -EFAULT;
|
2365 |
|
|
}
|
2366 |
|
|
|
2367 |
|
|
switch (cmd) {
|
2368 |
|
|
case ETHTOOL_GDRVINFO:
|
2369 |
|
|
if (copy_from_user(&info, addr, sizeof(info))) {
|
2370 |
|
|
return -EFAULT;
|
2371 |
|
|
}
|
2372 |
|
|
|
2373 |
|
|
if (strcmp(info.driver, "ifenslave") == 0) {
|
2374 |
|
|
int new_abi_ver;
|
2375 |
|
|
char *endptr;
|
2376 |
|
|
|
2377 |
|
|
new_abi_ver = simple_strtoul(info.fw_version,
|
2378 |
|
|
&endptr, 0);
|
2379 |
|
|
if (*endptr) {
|
2380 |
|
|
printk(KERN_ERR DRV_NAME
|
2381 |
|
|
": Error: got invalid ABI "
|
2382 |
|
|
"version from application\n");
|
2383 |
|
|
|
2384 |
|
|
return -EINVAL;
|
2385 |
|
|
}
|
2386 |
|
|
|
2387 |
|
|
if (orig_app_abi_ver == -1) {
|
2388 |
|
|
orig_app_abi_ver = new_abi_ver;
|
2389 |
|
|
}
|
2390 |
|
|
|
2391 |
|
|
app_abi_ver = new_abi_ver;
|
2392 |
|
|
}
|
2393 |
|
|
|
2394 |
|
|
strncpy(info.driver, DRV_NAME, 32);
|
2395 |
|
|
strncpy(info.version, DRV_VERSION, 32);
|
2396 |
|
|
snprintf(info.fw_version, 32, "%d", BOND_ABI_VERSION);
|
2397 |
|
|
|
2398 |
|
|
if (copy_to_user(addr, &info, sizeof(info))) {
|
2399 |
|
|
return -EFAULT;
|
2400 |
|
|
}
|
2401 |
|
|
|
2402 |
|
|
return 0;
|
2403 |
|
|
default:
|
2404 |
|
|
return -EOPNOTSUPP;
|
2405 |
|
|
}
|
2406 |
|
|
}
|
2407 |
|
|
|
2408 |
|
|
static int bond_info_query(struct net_device *bond_dev, struct ifbond *info)
|
2409 |
|
|
{
|
2410 |
|
|
struct bonding *bond = bond_dev->priv;
|
2411 |
|
|
|
2412 |
|
|
info->bond_mode = bond->params.mode;
|
2413 |
|
|
info->miimon = bond->params.miimon;
|
2414 |
|
|
|
2415 |
|
|
read_lock_bh(&bond->lock);
|
2416 |
|
|
info->num_slaves = bond->slave_cnt;
|
2417 |
|
|
read_unlock_bh(&bond->lock);
|
2418 |
|
|
|
2419 |
|
|
return 0;
|
2420 |
|
|
}
|
2421 |
|
|
|
2422 |
|
|
static int bond_slave_info_query(struct net_device *bond_dev, struct ifslave *info)
|
2423 |
|
|
{
|
2424 |
|
|
struct bonding *bond = bond_dev->priv;
|
2425 |
|
|
struct slave *slave;
|
2426 |
|
|
int i, found = 0;
|
2427 |
|
|
|
2428 |
|
|
if (info->slave_id < 0) {
|
2429 |
|
|
return -ENODEV;
|
2430 |
|
|
}
|
2431 |
|
|
|
2432 |
|
|
read_lock_bh(&bond->lock);
|
2433 |
|
|
|
2434 |
|
|
bond_for_each_slave(bond, slave, i) {
|
2435 |
|
|
if (i == (int)info->slave_id) {
|
2436 |
|
|
found = 1;
|
2437 |
|
|
break;
|
2438 |
|
|
}
|
2439 |
|
|
}
|
2440 |
|
|
|
2441 |
|
|
read_unlock_bh(&bond->lock);
|
2442 |
|
|
|
2443 |
|
|
if (found) {
|
2444 |
|
|
strcpy(info->slave_name, slave->dev->name);
|
2445 |
|
|
info->link = slave->link;
|
2446 |
|
|
info->state = slave->state;
|
2447 |
|
|
info->link_failure_count = slave->link_failure_count;
|
2448 |
|
|
} else {
|
2449 |
|
|
return -ENODEV;
|
2450 |
|
|
}
|
2451 |
|
|
|
2452 |
|
|
return 0;
|
2453 |
|
|
}
|
2454 |
|
|
|
2455 |
|
|
/*-------------------------------- Monitoring -------------------------------*/
|
2456 |
|
|
|
2457 |
|
|
/* this function is called regularly to monitor each slave's link. */
|
2458 |
|
|
static void bond_mii_monitor(struct net_device *bond_dev)
|
2459 |
|
|
{
|
2460 |
|
|
struct bonding *bond = bond_dev->priv;
|
2461 |
|
|
struct slave *slave, *oldcurrent;
|
2462 |
|
|
int do_failover = 0;
|
2463 |
|
|
int delta_in_ticks;
|
2464 |
|
|
int i;
|
2465 |
|
|
|
2466 |
|
|
read_lock(&bond->lock);
|
2467 |
|
|
|
2468 |
|
|
delta_in_ticks = (bond->params.miimon * HZ) / 1000;
|
2469 |
|
|
|
2470 |
|
|
if (bond->kill_timers) {
|
2471 |
|
|
goto out;
|
2472 |
|
|
}
|
2473 |
|
|
|
2474 |
|
|
if (bond->slave_cnt == 0) {
|
2475 |
|
|
goto re_arm;
|
2476 |
|
|
}
|
2477 |
|
|
|
2478 |
|
|
/* we will try to read the link status of each of our slaves, and
|
2479 |
|
|
* set their IFF_RUNNING flag appropriately. For each slave not
|
2480 |
|
|
* supporting MII status, we won't do anything so that a user-space
|
2481 |
|
|
* program could monitor the link itself if needed.
|
2482 |
|
|
*/
|
2483 |
|
|
|
2484 |
|
|
read_lock(&bond->curr_slave_lock);
|
2485 |
|
|
oldcurrent = bond->curr_active_slave;
|
2486 |
|
|
read_unlock(&bond->curr_slave_lock);
|
2487 |
|
|
|
2488 |
|
|
bond_for_each_slave(bond, slave, i) {
|
2489 |
|
|
struct net_device *slave_dev = slave->dev;
|
2490 |
|
|
int link_state;
|
2491 |
|
|
u16 old_speed = slave->speed;
|
2492 |
|
|
u8 old_duplex = slave->duplex;
|
2493 |
|
|
|
2494 |
|
|
link_state = bond_check_dev_link(bond, slave_dev, 0);
|
2495 |
|
|
|
2496 |
|
|
switch (slave->link) {
|
2497 |
|
|
case BOND_LINK_UP: /* the link was up */
|
2498 |
|
|
if (link_state == BMSR_LSTATUS) {
|
2499 |
|
|
/* link stays up, nothing more to do */
|
2500 |
|
|
break;
|
2501 |
|
|
} else { /* link going down */
|
2502 |
|
|
slave->link = BOND_LINK_FAIL;
|
2503 |
|
|
slave->delay = bond->params.downdelay;
|
2504 |
|
|
|
2505 |
|
|
if (slave->link_failure_count < UINT_MAX) {
|
2506 |
|
|
slave->link_failure_count++;
|
2507 |
|
|
}
|
2508 |
|
|
|
2509 |
|
|
if (bond->params.downdelay) {
|
2510 |
|
|
printk(KERN_INFO DRV_NAME
|
2511 |
|
|
": %s: link status down for %s "
|
2512 |
|
|
"interface %s, disabling it in "
|
2513 |
|
|
"%d ms.\n",
|
2514 |
|
|
bond_dev->name,
|
2515 |
|
|
IS_UP(slave_dev)
|
2516 |
|
|
? ((bond->params.mode == BOND_MODE_ACTIVEBACKUP)
|
2517 |
|
|
? ((slave == oldcurrent)
|
2518 |
|
|
? "active " : "backup ")
|
2519 |
|
|
: "")
|
2520 |
|
|
: "idle ",
|
2521 |
|
|
slave_dev->name,
|
2522 |
|
|
bond->params.downdelay * bond->params.miimon);
|
2523 |
|
|
}
|
2524 |
|
|
}
|
2525 |
|
|
/* no break ! fall through the BOND_LINK_FAIL test to
|
2526 |
|
|
ensure proper action to be taken
|
2527 |
|
|
*/
|
2528 |
|
|
case BOND_LINK_FAIL: /* the link has just gone down */
|
2529 |
|
|
if (link_state != BMSR_LSTATUS) {
|
2530 |
|
|
/* link stays down */
|
2531 |
|
|
if (slave->delay <= 0) {
|
2532 |
|
|
/* link down for too long time */
|
2533 |
|
|
slave->link = BOND_LINK_DOWN;
|
2534 |
|
|
|
2535 |
|
|
/* in active/backup mode, we must
|
2536 |
|
|
* completely disable this interface
|
2537 |
|
|
*/
|
2538 |
|
|
if ((bond->params.mode == BOND_MODE_ACTIVEBACKUP) ||
|
2539 |
|
|
(bond->params.mode == BOND_MODE_8023AD)) {
|
2540 |
|
|
bond_set_slave_inactive_flags(slave);
|
2541 |
|
|
}
|
2542 |
|
|
|
2543 |
|
|
printk(KERN_INFO DRV_NAME
|
2544 |
|
|
": %s: link status definitely "
|
2545 |
|
|
"down for interface %s, "
|
2546 |
|
|
"disabling it\n",
|
2547 |
|
|
bond_dev->name,
|
2548 |
|
|
slave_dev->name);
|
2549 |
|
|
|
2550 |
|
|
/* notify ad that the link status has changed */
|
2551 |
|
|
if (bond->params.mode == BOND_MODE_8023AD) {
|
2552 |
|
|
bond_3ad_handle_link_change(slave, BOND_LINK_DOWN);
|
2553 |
|
|
}
|
2554 |
|
|
|
2555 |
|
|
if ((bond->params.mode == BOND_MODE_TLB) ||
|
2556 |
|
|
(bond->params.mode == BOND_MODE_ALB)) {
|
2557 |
|
|
bond_alb_handle_link_change(bond, slave, BOND_LINK_DOWN);
|
2558 |
|
|
}
|
2559 |
|
|
|
2560 |
|
|
if (slave == oldcurrent) {
|
2561 |
|
|
do_failover = 1;
|
2562 |
|
|
}
|
2563 |
|
|
} else {
|
2564 |
|
|
slave->delay--;
|
2565 |
|
|
}
|
2566 |
|
|
} else {
|
2567 |
|
|
/* link up again */
|
2568 |
|
|
slave->link = BOND_LINK_UP;
|
2569 |
|
|
slave->jiffies = jiffies;
|
2570 |
|
|
printk(KERN_INFO DRV_NAME
|
2571 |
|
|
": %s: link status up again after %d "
|
2572 |
|
|
"ms for interface %s.\n",
|
2573 |
|
|
bond_dev->name,
|
2574 |
|
|
(bond->params.downdelay - slave->delay) * bond->params.miimon,
|
2575 |
|
|
slave_dev->name);
|
2576 |
|
|
}
|
2577 |
|
|
break;
|
2578 |
|
|
case BOND_LINK_DOWN: /* the link was down */
|
2579 |
|
|
if (link_state != BMSR_LSTATUS) {
|
2580 |
|
|
/* the link stays down, nothing more to do */
|
2581 |
|
|
break;
|
2582 |
|
|
} else { /* link going up */
|
2583 |
|
|
slave->link = BOND_LINK_BACK;
|
2584 |
|
|
slave->delay = bond->params.updelay;
|
2585 |
|
|
|
2586 |
|
|
if (bond->params.updelay) {
|
2587 |
|
|
/* if updelay == 0, no need to
|
2588 |
|
|
advertise about a 0 ms delay */
|
2589 |
|
|
printk(KERN_INFO DRV_NAME
|
2590 |
|
|
": %s: link status up for "
|
2591 |
|
|
"interface %s, enabling it "
|
2592 |
|
|
"in %d ms.\n",
|
2593 |
|
|
bond_dev->name,
|
2594 |
|
|
slave_dev->name,
|
2595 |
|
|
bond->params.updelay * bond->params.miimon);
|
2596 |
|
|
}
|
2597 |
|
|
}
|
2598 |
|
|
/* no break ! fall through the BOND_LINK_BACK state in
|
2599 |
|
|
case there's something to do.
|
2600 |
|
|
*/
|
2601 |
|
|
case BOND_LINK_BACK: /* the link has just come back */
|
2602 |
|
|
if (link_state != BMSR_LSTATUS) {
|
2603 |
|
|
/* link down again */
|
2604 |
|
|
slave->link = BOND_LINK_DOWN;
|
2605 |
|
|
|
2606 |
|
|
printk(KERN_INFO DRV_NAME
|
2607 |
|
|
": %s: link status down again after %d "
|
2608 |
|
|
"ms for interface %s.\n",
|
2609 |
|
|
bond_dev->name,
|
2610 |
|
|
(bond->params.updelay - slave->delay) * bond->params.miimon,
|
2611 |
|
|
slave_dev->name);
|
2612 |
|
|
} else {
|
2613 |
|
|
/* link stays up */
|
2614 |
|
|
if (slave->delay == 0) {
|
2615 |
|
|
/* now the link has been up for long time enough */
|
2616 |
|
|
slave->link = BOND_LINK_UP;
|
2617 |
|
|
slave->jiffies = jiffies;
|
2618 |
|
|
|
2619 |
|
|
if (bond->params.mode == BOND_MODE_8023AD) {
|
2620 |
|
|
/* prevent it from being the active one */
|
2621 |
|
|
slave->state = BOND_STATE_BACKUP;
|
2622 |
|
|
} else if (bond->params.mode != BOND_MODE_ACTIVEBACKUP) {
|
2623 |
|
|
/* make it immediately active */
|
2624 |
|
|
slave->state = BOND_STATE_ACTIVE;
|
2625 |
|
|
} else if (slave != bond->primary_slave) {
|
2626 |
|
|
/* prevent it from being the active one */
|
2627 |
|
|
slave->state = BOND_STATE_BACKUP;
|
2628 |
|
|
}
|
2629 |
|
|
|
2630 |
|
|
printk(KERN_INFO DRV_NAME
|
2631 |
|
|
": %s: link status definitely "
|
2632 |
|
|
"up for interface %s.\n",
|
2633 |
|
|
bond_dev->name,
|
2634 |
|
|
slave_dev->name);
|
2635 |
|
|
|
2636 |
|
|
/* notify ad that the link status has changed */
|
2637 |
|
|
if (bond->params.mode == BOND_MODE_8023AD) {
|
2638 |
|
|
bond_3ad_handle_link_change(slave, BOND_LINK_UP);
|
2639 |
|
|
}
|
2640 |
|
|
|
2641 |
|
|
if ((bond->params.mode == BOND_MODE_TLB) ||
|
2642 |
|
|
(bond->params.mode == BOND_MODE_ALB)) {
|
2643 |
|
|
bond_alb_handle_link_change(bond, slave, BOND_LINK_UP);
|
2644 |
|
|
}
|
2645 |
|
|
|
2646 |
|
|
if ((!oldcurrent) ||
|
2647 |
|
|
(slave == bond->primary_slave)) {
|
2648 |
|
|
do_failover = 1;
|
2649 |
|
|
}
|
2650 |
|
|
} else {
|
2651 |
|
|
slave->delay--;
|
2652 |
|
|
}
|
2653 |
|
|
}
|
2654 |
|
|
break;
|
2655 |
|
|
default:
|
2656 |
|
|
/* Should not happen */
|
2657 |
|
|
printk(KERN_ERR "bonding: Error: %s Illegal value (link=%d)\n",
|
2658 |
|
|
slave->dev->name, slave->link);
|
2659 |
|
|
goto out;
|
2660 |
|
|
} /* end of switch (slave->link) */
|
2661 |
|
|
|
2662 |
|
|
bond_update_speed_duplex(slave);
|
2663 |
|
|
|
2664 |
|
|
if (bond->params.mode == BOND_MODE_8023AD) {
|
2665 |
|
|
if (old_speed != slave->speed) {
|
2666 |
|
|
bond_3ad_adapter_speed_changed(slave);
|
2667 |
|
|
}
|
2668 |
|
|
|
2669 |
|
|
if (old_duplex != slave->duplex) {
|
2670 |
|
|
bond_3ad_adapter_duplex_changed(slave);
|
2671 |
|
|
}
|
2672 |
|
|
}
|
2673 |
|
|
|
2674 |
|
|
} /* end of for */
|
2675 |
|
|
|
2676 |
|
|
if (do_failover) {
|
2677 |
|
|
write_lock(&bond->curr_slave_lock);
|
2678 |
|
|
|
2679 |
|
|
bond_select_active_slave(bond);
|
2680 |
|
|
|
2681 |
|
|
if (oldcurrent && !bond->curr_active_slave) {
|
2682 |
|
|
printk(KERN_INFO DRV_NAME
|
2683 |
|
|
": %s: now running without any active "
|
2684 |
|
|
"interface !\n",
|
2685 |
|
|
bond_dev->name);
|
2686 |
|
|
}
|
2687 |
|
|
|
2688 |
|
|
write_unlock(&bond->curr_slave_lock);
|
2689 |
|
|
}
|
2690 |
|
|
|
2691 |
|
|
re_arm:
|
2692 |
|
|
if (bond->params.miimon) {
|
2693 |
|
|
mod_timer(&bond->mii_timer, jiffies + delta_in_ticks);
|
2694 |
|
|
}
|
2695 |
|
|
out:
|
2696 |
|
|
read_unlock(&bond->lock);
|
2697 |
|
|
}
|
2698 |
|
|
|
2699 |
|
|
static void bond_arp_send_all(struct bonding *bond, struct slave *slave)
|
2700 |
|
|
{
|
2701 |
|
|
int i;
|
2702 |
|
|
u32 *targets = bond->params.arp_targets;
|
2703 |
|
|
|
2704 |
|
|
for (i = 0; (i < BOND_MAX_ARP_TARGETS) && targets[i]; i++) {
|
2705 |
|
|
arp_send(ARPOP_REQUEST, ETH_P_ARP, targets[i], slave->dev,
|
2706 |
|
|
my_ip, NULL, slave->dev->dev_addr,
|
2707 |
|
|
NULL);
|
2708 |
|
|
}
|
2709 |
|
|
}
|
2710 |
|
|
|
2711 |
|
|
/*
|
2712 |
|
|
* this function is called regularly to monitor each slave's link
|
2713 |
|
|
* ensuring that traffic is being sent and received when arp monitoring
|
2714 |
|
|
* is used in load-balancing mode. if the adapter has been dormant, then an
|
2715 |
|
|
* arp is transmitted to generate traffic. see activebackup_arp_monitor for
|
2716 |
|
|
* arp monitoring in active backup mode.
|
2717 |
|
|
*/
|
2718 |
|
|
static void bond_loadbalance_arp_mon(struct net_device *bond_dev)
|
2719 |
|
|
{
|
2720 |
|
|
struct bonding *bond = bond_dev->priv;
|
2721 |
|
|
struct slave *slave, *oldcurrent;
|
2722 |
|
|
int do_failover = 0;
|
2723 |
|
|
int delta_in_ticks;
|
2724 |
|
|
int i;
|
2725 |
|
|
|
2726 |
|
|
read_lock(&bond->lock);
|
2727 |
|
|
|
2728 |
|
|
delta_in_ticks = (bond->params.arp_interval * HZ) / 1000;
|
2729 |
|
|
|
2730 |
|
|
if (bond->kill_timers) {
|
2731 |
|
|
goto out;
|
2732 |
|
|
}
|
2733 |
|
|
|
2734 |
|
|
if (bond->slave_cnt == 0) {
|
2735 |
|
|
goto re_arm;
|
2736 |
|
|
}
|
2737 |
|
|
|
2738 |
|
|
read_lock(&bond->curr_slave_lock);
|
2739 |
|
|
oldcurrent = bond->curr_active_slave;
|
2740 |
|
|
read_unlock(&bond->curr_slave_lock);
|
2741 |
|
|
|
2742 |
|
|
/* see if any of the previous devices are up now (i.e. they have
|
2743 |
|
|
* xmt and rcv traffic). the curr_active_slave does not come into
|
2744 |
|
|
* the picture unless it is null. also, slave->jiffies is not needed
|
2745 |
|
|
* here because we send an arp on each slave and give a slave as
|
2746 |
|
|
* long as it needs to get the tx/rx within the delta.
|
2747 |
|
|
* TODO: what about up/down delay in arp mode? it wasn't here before
|
2748 |
|
|
* so it can wait
|
2749 |
|
|
*/
|
2750 |
|
|
bond_for_each_slave(bond, slave, i) {
|
2751 |
|
|
if (slave->link != BOND_LINK_UP) {
|
2752 |
|
|
if (((jiffies - slave->dev->trans_start) <= delta_in_ticks) &&
|
2753 |
|
|
((jiffies - slave->dev->last_rx) <= delta_in_ticks)) {
|
2754 |
|
|
|
2755 |
|
|
slave->link = BOND_LINK_UP;
|
2756 |
|
|
slave->state = BOND_STATE_ACTIVE;
|
2757 |
|
|
|
2758 |
|
|
/* primary_slave has no meaning in round-robin
|
2759 |
|
|
* mode. the window of a slave being up and
|
2760 |
|
|
* curr_active_slave being null after enslaving
|
2761 |
|
|
* is closed.
|
2762 |
|
|
*/
|
2763 |
|
|
if (!oldcurrent) {
|
2764 |
|
|
printk(KERN_INFO DRV_NAME
|
2765 |
|
|
": %s: link status definitely "
|
2766 |
|
|
"up for interface %s, ",
|
2767 |
|
|
bond_dev->name,
|
2768 |
|
|
slave->dev->name);
|
2769 |
|
|
do_failover = 1;
|
2770 |
|
|
} else {
|
2771 |
|
|
printk(KERN_INFO DRV_NAME
|
2772 |
|
|
": %s: interface %s is now up\n",
|
2773 |
|
|
bond_dev->name,
|
2774 |
|
|
slave->dev->name);
|
2775 |
|
|
}
|
2776 |
|
|
}
|
2777 |
|
|
} else {
|
2778 |
|
|
/* slave->link == BOND_LINK_UP */
|
2779 |
|
|
|
2780 |
|
|
/* not all switches will respond to an arp request
|
2781 |
|
|
* when the source ip is 0, so don't take the link down
|
2782 |
|
|
* if we don't know our ip yet
|
2783 |
|
|
*/
|
2784 |
|
|
if (((jiffies - slave->dev->trans_start) >= (2*delta_in_ticks)) ||
|
2785 |
|
|
(((jiffies - slave->dev->last_rx) >= (2*delta_in_ticks)) &&
|
2786 |
|
|
my_ip)) {
|
2787 |
|
|
|
2788 |
|
|
slave->link = BOND_LINK_DOWN;
|
2789 |
|
|
slave->state = BOND_STATE_BACKUP;
|
2790 |
|
|
|
2791 |
|
|
if (slave->link_failure_count < UINT_MAX) {
|
2792 |
|
|
slave->link_failure_count++;
|
2793 |
|
|
}
|
2794 |
|
|
|
2795 |
|
|
printk(KERN_INFO DRV_NAME
|
2796 |
|
|
": %s: interface %s is now down.\n",
|
2797 |
|
|
bond_dev->name,
|
2798 |
|
|
slave->dev->name);
|
2799 |
|
|
|
2800 |
|
|
if (slave == oldcurrent) {
|
2801 |
|
|
do_failover = 1;
|
2802 |
|
|
}
|
2803 |
|
|
}
|
2804 |
|
|
}
|
2805 |
|
|
|
2806 |
|
|
/* note: if switch is in round-robin mode, all links
|
2807 |
|
|
* must tx arp to ensure all links rx an arp - otherwise
|
2808 |
|
|
* links may oscillate or not come up at all; if switch is
|
2809 |
|
|
* in something like xor mode, there is nothing we can
|
2810 |
|
|
* do - all replies will be rx'ed on same link causing slaves
|
2811 |
|
|
* to be unstable during low/no traffic periods
|
2812 |
|
|
*/
|
2813 |
|
|
if (IS_UP(slave->dev)) {
|
2814 |
|
|
bond_arp_send_all(bond, slave);
|
2815 |
|
|
}
|
2816 |
|
|
}
|
2817 |
|
|
|
2818 |
|
|
if (do_failover) {
|
2819 |
|
|
write_lock(&bond->curr_slave_lock);
|
2820 |
|
|
|
2821 |
|
|
bond_select_active_slave(bond);
|
2822 |
|
|
|
2823 |
|
|
if (oldcurrent && !bond->curr_active_slave) {
|
2824 |
|
|
printk(KERN_INFO DRV_NAME
|
2825 |
|
|
": %s: now running without any active "
|
2826 |
|
|
"interface !\n",
|
2827 |
|
|
bond_dev->name);
|
2828 |
|
|
}
|
2829 |
|
|
|
2830 |
|
|
write_unlock(&bond->curr_slave_lock);
|
2831 |
|
|
}
|
2832 |
|
|
|
2833 |
|
|
re_arm:
|
2834 |
|
|
if (bond->params.arp_interval) {
|
2835 |
|
|
mod_timer(&bond->arp_timer, jiffies + delta_in_ticks);
|
2836 |
|
|
}
|
2837 |
|
|
out:
|
2838 |
|
|
read_unlock(&bond->lock);
|
2839 |
|
|
}
|
2840 |
|
|
|
2841 |
|
|
/*
|
2842 |
|
|
* When using arp monitoring in active-backup mode, this function is
|
2843 |
|
|
* called to determine if any backup slaves have went down or a new
|
2844 |
|
|
* current slave needs to be found.
|
2845 |
|
|
* The backup slaves never generate traffic, they are considered up by merely
|
2846 |
|
|
* receiving traffic. If the current slave goes down, each backup slave will
|
2847 |
|
|
* be given the opportunity to tx/rx an arp before being taken down - this
|
2848 |
|
|
* prevents all slaves from being taken down due to the current slave not
|
2849 |
|
|
* sending any traffic for the backups to receive. The arps are not necessarily
|
2850 |
|
|
* necessary, any tx and rx traffic will keep the current slave up. While any
|
2851 |
|
|
* rx traffic will keep the backup slaves up, the current slave is responsible
|
2852 |
|
|
* for generating traffic to keep them up regardless of any other traffic they
|
2853 |
|
|
* may have received.
|
2854 |
|
|
* see loadbalance_arp_monitor for arp monitoring in load balancing mode
|
2855 |
|
|
*/
|
2856 |
|
|
static void bond_activebackup_arp_mon(struct net_device *bond_dev)
|
2857 |
|
|
{
|
2858 |
|
|
struct bonding *bond = bond_dev->priv;
|
2859 |
|
|
struct slave *slave;
|
2860 |
|
|
int delta_in_ticks;
|
2861 |
|
|
int i;
|
2862 |
|
|
|
2863 |
|
|
read_lock(&bond->lock);
|
2864 |
|
|
|
2865 |
|
|
delta_in_ticks = (bond->params.arp_interval * HZ) / 1000;
|
2866 |
|
|
|
2867 |
|
|
if (bond->kill_timers) {
|
2868 |
|
|
goto out;
|
2869 |
|
|
}
|
2870 |
|
|
|
2871 |
|
|
if (bond->slave_cnt == 0) {
|
2872 |
|
|
goto re_arm;
|
2873 |
|
|
}
|
2874 |
|
|
|
2875 |
|
|
/* determine if any slave has come up or any backup slave has
|
2876 |
|
|
* gone down
|
2877 |
|
|
* TODO: what about up/down delay in arp mode? it wasn't here before
|
2878 |
|
|
* so it can wait
|
2879 |
|
|
*/
|
2880 |
|
|
bond_for_each_slave(bond, slave, i) {
|
2881 |
|
|
if (slave->link != BOND_LINK_UP) {
|
2882 |
|
|
if ((jiffies - slave->dev->last_rx) <= delta_in_ticks) {
|
2883 |
|
|
|
2884 |
|
|
slave->link = BOND_LINK_UP;
|
2885 |
|
|
|
2886 |
|
|
write_lock(&bond->curr_slave_lock);
|
2887 |
|
|
|
2888 |
|
|
if ((!bond->curr_active_slave) &&
|
2889 |
|
|
((jiffies - slave->dev->trans_start) <= delta_in_ticks)) {
|
2890 |
|
|
bond_change_active_slave(bond, slave);
|
2891 |
|
|
bond->current_arp_slave = NULL;
|
2892 |
|
|
} else if (bond->curr_active_slave != slave) {
|
2893 |
|
|
/* this slave has just come up but we
|
2894 |
|
|
* already have a current slave; this
|
2895 |
|
|
* can also happen if bond_enslave adds
|
2896 |
|
|
* a new slave that is up while we are
|
2897 |
|
|
* searching for a new slave
|
2898 |
|
|
*/
|
2899 |
|
|
bond_set_slave_inactive_flags(slave);
|
2900 |
|
|
bond->current_arp_slave = NULL;
|
2901 |
|
|
}
|
2902 |
|
|
|
2903 |
|
|
if (slave == bond->curr_active_slave) {
|
2904 |
|
|
printk(KERN_INFO DRV_NAME
|
2905 |
|
|
": %s: %s is up and now the "
|
2906 |
|
|
"active interface\n",
|
2907 |
|
|
bond_dev->name,
|
2908 |
|
|
slave->dev->name);
|
2909 |
|
|
} else {
|
2910 |
|
|
printk(KERN_INFO DRV_NAME
|
2911 |
|
|
": %s: backup interface %s is "
|
2912 |
|
|
"now up\n",
|
2913 |
|
|
bond_dev->name,
|
2914 |
|
|
slave->dev->name);
|
2915 |
|
|
}
|
2916 |
|
|
|
2917 |
|
|
write_unlock(&bond->curr_slave_lock);
|
2918 |
|
|
}
|
2919 |
|
|
} else {
|
2920 |
|
|
read_lock(&bond->curr_slave_lock);
|
2921 |
|
|
|
2922 |
|
|
if ((slave != bond->curr_active_slave) &&
|
2923 |
|
|
(!bond->current_arp_slave) &&
|
2924 |
|
|
(((jiffies - slave->dev->last_rx) >= 3*delta_in_ticks) &&
|
2925 |
|
|
my_ip)) {
|
2926 |
|
|
/* a backup slave has gone down; three times
|
2927 |
|
|
* the delta allows the current slave to be
|
2928 |
|
|
* taken out before the backup slave.
|
2929 |
|
|
* note: a non-null current_arp_slave indicates
|
2930 |
|
|
* the curr_active_slave went down and we are
|
2931 |
|
|
* searching for a new one; under this
|
2932 |
|
|
* condition we only take the curr_active_slave
|
2933 |
|
|
* down - this gives each slave a chance to
|
2934 |
|
|
* tx/rx traffic before being taken out
|
2935 |
|
|
*/
|
2936 |
|
|
|
2937 |
|
|
read_unlock(&bond->curr_slave_lock);
|
2938 |
|
|
|
2939 |
|
|
slave->link = BOND_LINK_DOWN;
|
2940 |
|
|
|
2941 |
|
|
if (slave->link_failure_count < UINT_MAX) {
|
2942 |
|
|
slave->link_failure_count++;
|
2943 |
|
|
}
|
2944 |
|
|
|
2945 |
|
|
bond_set_slave_inactive_flags(slave);
|
2946 |
|
|
|
2947 |
|
|
printk(KERN_INFO DRV_NAME
|
2948 |
|
|
": %s: backup interface %s is now down\n",
|
2949 |
|
|
bond_dev->name,
|
2950 |
|
|
slave->dev->name);
|
2951 |
|
|
} else {
|
2952 |
|
|
read_unlock(&bond->curr_slave_lock);
|
2953 |
|
|
}
|
2954 |
|
|
}
|
2955 |
|
|
}
|
2956 |
|
|
|
2957 |
|
|
read_lock(&bond->curr_slave_lock);
|
2958 |
|
|
slave = bond->curr_active_slave;
|
2959 |
|
|
read_unlock(&bond->curr_slave_lock);
|
2960 |
|
|
|
2961 |
|
|
if (slave) {
|
2962 |
|
|
/* if we have sent traffic in the past 2*arp_intervals but
|
2963 |
|
|
* haven't xmit and rx traffic in that time interval, select
|
2964 |
|
|
* a different slave. slave->jiffies is only updated when
|
2965 |
|
|
* a slave first becomes the curr_active_slave - not necessarily
|
2966 |
|
|
* after every arp; this ensures the slave has a full 2*delta
|
2967 |
|
|
* before being taken out. if a primary is being used, check
|
2968 |
|
|
* if it is up and needs to take over as the curr_active_slave
|
2969 |
|
|
*/
|
2970 |
|
|
if ((((jiffies - slave->dev->trans_start) >= (2*delta_in_ticks)) ||
|
2971 |
|
|
(((jiffies - slave->dev->last_rx) >= (2*delta_in_ticks)) &&
|
2972 |
|
|
my_ip)) &&
|
2973 |
|
|
((jiffies - slave->jiffies) >= 2*delta_in_ticks)) {
|
2974 |
|
|
|
2975 |
|
|
slave->link = BOND_LINK_DOWN;
|
2976 |
|
|
|
2977 |
|
|
if (slave->link_failure_count < UINT_MAX) {
|
2978 |
|
|
slave->link_failure_count++;
|
2979 |
|
|
}
|
2980 |
|
|
|
2981 |
|
|
printk(KERN_INFO DRV_NAME
|
2982 |
|
|
": %s: link status down for active interface "
|
2983 |
|
|
"%s, disabling it\n",
|
2984 |
|
|
bond_dev->name,
|
2985 |
|
|
slave->dev->name);
|
2986 |
|
|
|
2987 |
|
|
write_lock(&bond->curr_slave_lock);
|
2988 |
|
|
|
2989 |
|
|
bond_select_active_slave(bond);
|
2990 |
|
|
slave = bond->curr_active_slave;
|
2991 |
|
|
|
2992 |
|
|
write_unlock(&bond->curr_slave_lock);
|
2993 |
|
|
|
2994 |
|
|
bond->current_arp_slave = slave;
|
2995 |
|
|
|
2996 |
|
|
if (slave) {
|
2997 |
|
|
slave->jiffies = jiffies;
|
2998 |
|
|
}
|
2999 |
|
|
} else if ((bond->primary_slave) &&
|
3000 |
|
|
(bond->primary_slave != slave) &&
|
3001 |
|
|
(bond->primary_slave->link == BOND_LINK_UP)) {
|
3002 |
|
|
/* at this point, slave is the curr_active_slave */
|
3003 |
|
|
printk(KERN_INFO DRV_NAME
|
3004 |
|
|
": %s: changing from interface %s to primary "
|
3005 |
|
|
"interface %s\n",
|
3006 |
|
|
bond_dev->name,
|
3007 |
|
|
slave->dev->name,
|
3008 |
|
|
bond->primary_slave->dev->name);
|
3009 |
|
|
|
3010 |
|
|
/* primary is up so switch to it */
|
3011 |
|
|
write_lock(&bond->curr_slave_lock);
|
3012 |
|
|
bond_change_active_slave(bond, bond->primary_slave);
|
3013 |
|
|
write_unlock(&bond->curr_slave_lock);
|
3014 |
|
|
|
3015 |
|
|
slave = bond->primary_slave;
|
3016 |
|
|
slave->jiffies = jiffies;
|
3017 |
|
|
} else {
|
3018 |
|
|
bond->current_arp_slave = NULL;
|
3019 |
|
|
}
|
3020 |
|
|
|
3021 |
|
|
/* the current slave must tx an arp to ensure backup slaves
|
3022 |
|
|
* rx traffic
|
3023 |
|
|
*/
|
3024 |
|
|
if (slave && my_ip) {
|
3025 |
|
|
bond_arp_send_all(bond, slave);
|
3026 |
|
|
}
|
3027 |
|
|
}
|
3028 |
|
|
|
3029 |
|
|
/* if we don't have a curr_active_slave, search for the next available
|
3030 |
|
|
* backup slave from the current_arp_slave and make it the candidate
|
3031 |
|
|
* for becoming the curr_active_slave
|
3032 |
|
|
*/
|
3033 |
|
|
if (!slave) {
|
3034 |
|
|
if (!bond->current_arp_slave) {
|
3035 |
|
|
bond->current_arp_slave = bond->first_slave;
|
3036 |
|
|
}
|
3037 |
|
|
|
3038 |
|
|
if (bond->current_arp_slave) {
|
3039 |
|
|
bond_set_slave_inactive_flags(bond->current_arp_slave);
|
3040 |
|
|
|
3041 |
|
|
/* search for next candidate */
|
3042 |
|
|
bond_for_each_slave_from(bond, slave, i, bond->current_arp_slave) {
|
3043 |
|
|
if (IS_UP(slave->dev)) {
|
3044 |
|
|
slave->link = BOND_LINK_BACK;
|
3045 |
|
|
bond_set_slave_active_flags(slave);
|
3046 |
|
|
bond_arp_send_all(bond, slave);
|
3047 |
|
|
slave->jiffies = jiffies;
|
3048 |
|
|
bond->current_arp_slave = slave;
|
3049 |
|
|
break;
|
3050 |
|
|
}
|
3051 |
|
|
|
3052 |
|
|
/* if the link state is up at this point, we
|
3053 |
|
|
* mark it down - this can happen if we have
|
3054 |
|
|
* simultaneous link failures and
|
3055 |
|
|
* reselect_active_interface doesn't make this
|
3056 |
|
|
* one the current slave so it is still marked
|
3057 |
|
|
* up when it is actually down
|
3058 |
|
|
*/
|
3059 |
|
|
if (slave->link == BOND_LINK_UP) {
|
3060 |
|
|
slave->link = BOND_LINK_DOWN;
|
3061 |
|
|
if (slave->link_failure_count < UINT_MAX) {
|
3062 |
|
|
slave->link_failure_count++;
|
3063 |
|
|
}
|
3064 |
|
|
|
3065 |
|
|
bond_set_slave_inactive_flags(slave);
|
3066 |
|
|
|
3067 |
|
|
printk(KERN_INFO DRV_NAME
|
3068 |
|
|
": %s: backup interface %s is "
|
3069 |
|
|
"now down.\n",
|
3070 |
|
|
bond_dev->name,
|
3071 |
|
|
slave->dev->name);
|
3072 |
|
|
}
|
3073 |
|
|
}
|
3074 |
|
|
}
|
3075 |
|
|
}
|
3076 |
|
|
|
3077 |
|
|
re_arm:
|
3078 |
|
|
if (bond->params.arp_interval) {
|
3079 |
|
|
mod_timer(&bond->arp_timer, jiffies + delta_in_ticks);
|
3080 |
|
|
}
|
3081 |
|
|
out:
|
3082 |
|
|
read_unlock(&bond->lock);
|
3083 |
|
|
}
|
3084 |
|
|
|
3085 |
|
|
/*------------------------------ proc/seq_file-------------------------------*/
|
3086 |
|
|
|
3087 |
|
|
#ifdef CONFIG_PROC_FS
|
3088 |
|
|
|
3089 |
|
|
#define SEQ_START_TOKEN ((void *)1)
|
3090 |
|
|
|
3091 |
|
|
static void *bond_info_seq_start(struct seq_file *seq, loff_t *pos)
|
3092 |
|
|
{
|
3093 |
|
|
struct bonding *bond = seq->private;
|
3094 |
|
|
loff_t off = 0;
|
3095 |
|
|
struct slave *slave;
|
3096 |
|
|
int i;
|
3097 |
|
|
|
3098 |
|
|
/* make sure the bond won't be taken away */
|
3099 |
|
|
read_lock(&dev_base_lock);
|
3100 |
|
|
read_lock_bh(&bond->lock);
|
3101 |
|
|
|
3102 |
|
|
if (*pos == 0) {
|
3103 |
|
|
return SEQ_START_TOKEN;
|
3104 |
|
|
}
|
3105 |
|
|
|
3106 |
|
|
bond_for_each_slave(bond, slave, i) {
|
3107 |
|
|
if (++off == *pos) {
|
3108 |
|
|
return slave;
|
3109 |
|
|
}
|
3110 |
|
|
}
|
3111 |
|
|
|
3112 |
|
|
return NULL;
|
3113 |
|
|
}
|
3114 |
|
|
|
3115 |
|
|
static void *bond_info_seq_next(struct seq_file *seq, void *v, loff_t *pos)
|
3116 |
|
|
{
|
3117 |
|
|
struct bonding *bond = seq->private;
|
3118 |
|
|
struct slave *slave = v;
|
3119 |
|
|
|
3120 |
|
|
++*pos;
|
3121 |
|
|
if (v == SEQ_START_TOKEN) {
|
3122 |
|
|
return bond->first_slave;
|
3123 |
|
|
}
|
3124 |
|
|
|
3125 |
|
|
slave = slave->next;
|
3126 |
|
|
|
3127 |
|
|
return (slave == bond->first_slave) ? NULL : slave;
|
3128 |
|
|
}
|
3129 |
|
|
|
3130 |
|
|
static void bond_info_seq_stop(struct seq_file *seq, void *v)
|
3131 |
|
|
{
|
3132 |
|
|
struct bonding *bond = seq->private;
|
3133 |
|
|
|
3134 |
|
|
read_unlock_bh(&bond->lock);
|
3135 |
|
|
read_unlock(&dev_base_lock);
|
3136 |
|
|
}
|
3137 |
|
|
|
3138 |
|
|
static void bond_info_show_master(struct seq_file *seq)
|
3139 |
|
|
{
|
3140 |
|
|
struct bonding *bond = seq->private;
|
3141 |
|
|
struct slave *curr;
|
3142 |
|
|
|
3143 |
|
|
read_lock(&bond->curr_slave_lock);
|
3144 |
|
|
curr = bond->curr_active_slave;
|
3145 |
|
|
read_unlock(&bond->curr_slave_lock);
|
3146 |
|
|
|
3147 |
|
|
seq_printf(seq, "Bonding Mode: %s\n",
|
3148 |
|
|
bond_mode_name(bond->params.mode));
|
3149 |
|
|
|
3150 |
|
|
if (USES_PRIMARY(bond->params.mode)) {
|
3151 |
|
|
seq_printf(seq, "Primary Slave: %s\n",
|
3152 |
|
|
(bond->params.primary[0]) ?
|
3153 |
|
|
bond->params.primary : "None");
|
3154 |
|
|
|
3155 |
|
|
seq_printf(seq, "Currently Active Slave: %s\n",
|
3156 |
|
|
(curr) ? curr->dev->name : "None");
|
3157 |
|
|
}
|
3158 |
|
|
|
3159 |
|
|
seq_printf(seq, "MII Status: %s\n", (curr) ? "up" : "down");
|
3160 |
|
|
seq_printf(seq, "MII Polling Interval (ms): %d\n", bond->params.miimon);
|
3161 |
|
|
seq_printf(seq, "Up Delay (ms): %d\n",
|
3162 |
|
|
bond->params.updelay * bond->params.miimon);
|
3163 |
|
|
seq_printf(seq, "Down Delay (ms): %d\n",
|
3164 |
|
|
bond->params.downdelay * bond->params.miimon);
|
3165 |
|
|
|
3166 |
|
|
if (bond->params.mode == BOND_MODE_8023AD) {
|
3167 |
|
|
struct ad_info ad_info;
|
3168 |
|
|
|
3169 |
|
|
seq_puts(seq, "\n802.3ad info\n");
|
3170 |
|
|
seq_printf(seq, "LACP rate: %s\n",
|
3171 |
|
|
(bond->params.lacp_fast) ? "fast" : "slow");
|
3172 |
|
|
|
3173 |
|
|
if (bond_3ad_get_active_agg_info(bond, &ad_info)) {
|
3174 |
|
|
seq_printf(seq, "bond %s has no active aggregator\n",
|
3175 |
|
|
bond->dev->name);
|
3176 |
|
|
} else {
|
3177 |
|
|
seq_printf(seq, "Active Aggregator Info:\n");
|
3178 |
|
|
|
3179 |
|
|
seq_printf(seq, "\tAggregator ID: %d\n",
|
3180 |
|
|
ad_info.aggregator_id);
|
3181 |
|
|
seq_printf(seq, "\tNumber of ports: %d\n",
|
3182 |
|
|
ad_info.ports);
|
3183 |
|
|
seq_printf(seq, "\tActor Key: %d\n",
|
3184 |
|
|
ad_info.actor_key);
|
3185 |
|
|
seq_printf(seq, "\tPartner Key: %d\n",
|
3186 |
|
|
ad_info.partner_key);
|
3187 |
|
|
seq_printf(seq, "\tPartner Mac Address: %02x:%02x:%02x:%02x:%02x:%02x\n",
|
3188 |
|
|
ad_info.partner_system[0],
|
3189 |
|
|
ad_info.partner_system[1],
|
3190 |
|
|
ad_info.partner_system[2],
|
3191 |
|
|
ad_info.partner_system[3],
|
3192 |
|
|
ad_info.partner_system[4],
|
3193 |
|
|
ad_info.partner_system[5]);
|
3194 |
|
|
}
|
3195 |
|
|
}
|
3196 |
|
|
}
|
3197 |
|
|
|
3198 |
|
|
static void bond_info_show_slave(struct seq_file *seq, const struct slave *slave)
|
3199 |
|
|
{
|
3200 |
|
|
struct bonding *bond = seq->private;
|
3201 |
|
|
|
3202 |
|
|
seq_printf(seq, "\nSlave Interface: %s\n", slave->dev->name);
|
3203 |
|
|
seq_printf(seq, "MII Status: %s\n",
|
3204 |
|
|
(slave->link == BOND_LINK_UP) ? "up" : "down");
|
3205 |
|
|
seq_printf(seq, "Link Failure Count: %d\n",
|
3206 |
|
|
slave->link_failure_count);
|
3207 |
|
|
|
3208 |
|
|
if (app_abi_ver >= 1) {
|
3209 |
|
|
seq_printf(seq,
|
3210 |
|
|
"Permanent HW addr: %02x:%02x:%02x:%02x:%02x:%02x\n",
|
3211 |
|
|
slave->perm_hwaddr[0],
|
3212 |
|
|
slave->perm_hwaddr[1],
|
3213 |
|
|
slave->perm_hwaddr[2],
|
3214 |
|
|
slave->perm_hwaddr[3],
|
3215 |
|
|
slave->perm_hwaddr[4],
|
3216 |
|
|
slave->perm_hwaddr[5]);
|
3217 |
|
|
}
|
3218 |
|
|
|
3219 |
|
|
if (bond->params.mode == BOND_MODE_8023AD) {
|
3220 |
|
|
const struct aggregator *agg
|
3221 |
|
|
= SLAVE_AD_INFO(slave).port.aggregator;
|
3222 |
|
|
|
3223 |
|
|
if (agg) {
|
3224 |
|
|
seq_printf(seq, "Aggregator ID: %d\n",
|
3225 |
|
|
agg->aggregator_identifier);
|
3226 |
|
|
} else {
|
3227 |
|
|
seq_puts(seq, "Aggregator ID: N/A\n");
|
3228 |
|
|
}
|
3229 |
|
|
}
|
3230 |
|
|
}
|
3231 |
|
|
|
3232 |
|
|
static int bond_info_seq_show(struct seq_file *seq, void *v)
|
3233 |
|
|
{
|
3234 |
|
|
if (v == SEQ_START_TOKEN) {
|
3235 |
|
|
seq_printf(seq, "%s\n", version);
|
3236 |
|
|
bond_info_show_master(seq);
|
3237 |
|
|
} else {
|
3238 |
|
|
bond_info_show_slave(seq, v);
|
3239 |
|
|
}
|
3240 |
|
|
|
3241 |
|
|
return 0;
|
3242 |
|
|
}
|
3243 |
|
|
|
3244 |
|
|
static struct seq_operations bond_info_seq_ops = {
|
3245 |
|
|
.start = bond_info_seq_start,
|
3246 |
|
|
.next = bond_info_seq_next,
|
3247 |
|
|
.stop = bond_info_seq_stop,
|
3248 |
|
|
.show = bond_info_seq_show,
|
3249 |
|
|
};
|
3250 |
|
|
|
3251 |
|
|
static int bond_info_open(struct inode *inode, struct file *file)
|
3252 |
|
|
{
|
3253 |
|
|
struct seq_file *seq;
|
3254 |
|
|
struct proc_dir_entry *proc;
|
3255 |
|
|
int res;
|
3256 |
|
|
|
3257 |
|
|
res = seq_open(file, &bond_info_seq_ops);
|
3258 |
|
|
if (!res) {
|
3259 |
|
|
/* recover the pointer buried in proc_dir_entry data */
|
3260 |
|
|
seq = file->private_data;
|
3261 |
|
|
proc = PDE(inode);
|
3262 |
|
|
seq->private = proc->data;
|
3263 |
|
|
}
|
3264 |
|
|
|
3265 |
|
|
return res;
|
3266 |
|
|
}
|
3267 |
|
|
|
3268 |
|
|
static struct file_operations bond_info_fops = {
|
3269 |
|
|
.owner = THIS_MODULE,
|
3270 |
|
|
.open = bond_info_open,
|
3271 |
|
|
.read = seq_read,
|
3272 |
|
|
.llseek = seq_lseek,
|
3273 |
|
|
.release = seq_release,
|
3274 |
|
|
};
|
3275 |
|
|
|
3276 |
|
|
static int bond_create_proc_entry(struct bonding *bond)
|
3277 |
|
|
{
|
3278 |
|
|
struct net_device *bond_dev = bond->dev;
|
3279 |
|
|
|
3280 |
|
|
if (bond_proc_dir) {
|
3281 |
|
|
bond->proc_entry = create_proc_entry(bond_dev->name,
|
3282 |
|
|
S_IRUGO,
|
3283 |
|
|
bond_proc_dir);
|
3284 |
|
|
if (bond->proc_entry == NULL) {
|
3285 |
|
|
printk(KERN_WARNING DRV_NAME
|
3286 |
|
|
": Warning: Cannot create /proc/net/%s/%s\n",
|
3287 |
|
|
DRV_NAME, bond_dev->name);
|
3288 |
|
|
} else {
|
3289 |
|
|
bond->proc_entry->data = bond;
|
3290 |
|
|
bond->proc_entry->proc_fops = &bond_info_fops;
|
3291 |
|
|
bond->proc_entry->owner = THIS_MODULE;
|
3292 |
|
|
memcpy(bond->proc_file_name, bond_dev->name, IFNAMSIZ);
|
3293 |
|
|
}
|
3294 |
|
|
}
|
3295 |
|
|
|
3296 |
|
|
return 0;
|
3297 |
|
|
}
|
3298 |
|
|
|
3299 |
|
|
static void bond_remove_proc_entry(struct bonding *bond)
|
3300 |
|
|
{
|
3301 |
|
|
if (bond_proc_dir && bond->proc_entry) {
|
3302 |
|
|
remove_proc_entry(bond->proc_file_name, bond_proc_dir);
|
3303 |
|
|
memset(bond->proc_file_name, 0, IFNAMSIZ);
|
3304 |
|
|
bond->proc_entry = NULL;
|
3305 |
|
|
}
|
3306 |
|
|
}
|
3307 |
|
|
|
3308 |
|
|
/* Create the bonding directory under /proc/net, if doesn't exist yet.
|
3309 |
|
|
* Caller must hold rtnl_lock.
|
3310 |
|
|
*/
|
3311 |
|
|
static void bond_create_proc_dir(void)
|
3312 |
|
|
{
|
3313 |
|
|
int len = strlen(DRV_NAME);
|
3314 |
|
|
|
3315 |
|
|
for (bond_proc_dir = proc_net->subdir; bond_proc_dir;
|
3316 |
|
|
bond_proc_dir = bond_proc_dir->next) {
|
3317 |
|
|
if ((bond_proc_dir->namelen == len) &&
|
3318 |
|
|
!memcmp(bond_proc_dir->name, DRV_NAME, len)) {
|
3319 |
|
|
break;
|
3320 |
|
|
}
|
3321 |
|
|
}
|
3322 |
|
|
|
3323 |
|
|
if (!bond_proc_dir) {
|
3324 |
|
|
bond_proc_dir = proc_mkdir(DRV_NAME, proc_net);
|
3325 |
|
|
if (bond_proc_dir) {
|
3326 |
|
|
bond_proc_dir->owner = THIS_MODULE;
|
3327 |
|
|
} else {
|
3328 |
|
|
printk(KERN_WARNING DRV_NAME
|
3329 |
|
|
": Warning: cannot create /proc/net/%s\n",
|
3330 |
|
|
DRV_NAME);
|
3331 |
|
|
}
|
3332 |
|
|
}
|
3333 |
|
|
}
|
3334 |
|
|
|
3335 |
|
|
/* Destroy the bonding directory under /proc/net, if empty.
|
3336 |
|
|
* Caller must hold rtnl_lock.
|
3337 |
|
|
*/
|
3338 |
|
|
static void bond_destroy_proc_dir(void)
|
3339 |
|
|
{
|
3340 |
|
|
struct proc_dir_entry *de;
|
3341 |
|
|
|
3342 |
|
|
if (!bond_proc_dir) {
|
3343 |
|
|
return;
|
3344 |
|
|
}
|
3345 |
|
|
|
3346 |
|
|
/* verify that the /proc dir is empty */
|
3347 |
|
|
for (de = bond_proc_dir->subdir; de; de = de->next) {
|
3348 |
|
|
/* ignore . and .. */
|
3349 |
|
|
if (*(de->name) != '.') {
|
3350 |
|
|
break;
|
3351 |
|
|
}
|
3352 |
|
|
}
|
3353 |
|
|
|
3354 |
|
|
if (de) {
|
3355 |
|
|
if (bond_proc_dir->owner == THIS_MODULE) {
|
3356 |
|
|
bond_proc_dir->owner = NULL;
|
3357 |
|
|
}
|
3358 |
|
|
} else {
|
3359 |
|
|
remove_proc_entry(DRV_NAME, proc_net);
|
3360 |
|
|
bond_proc_dir = NULL;
|
3361 |
|
|
}
|
3362 |
|
|
}
|
3363 |
|
|
#endif /* CONFIG_PROC_FS */
|
3364 |
|
|
|
3365 |
|
|
/*-------------------------- netdev event handling --------------------------*/
|
3366 |
|
|
|
3367 |
|
|
/*
|
3368 |
|
|
* Change device name
|
3369 |
|
|
*/
|
3370 |
|
|
static int bond_event_changename(struct bonding *bond)
|
3371 |
|
|
{
|
3372 |
|
|
#ifdef CONFIG_PROC_FS
|
3373 |
|
|
bond_remove_proc_entry(bond);
|
3374 |
|
|
bond_create_proc_entry(bond);
|
3375 |
|
|
#endif
|
3376 |
|
|
|
3377 |
|
|
return NOTIFY_DONE;
|
3378 |
|
|
}
|
3379 |
|
|
|
3380 |
|
|
static int bond_master_netdev_event(unsigned long event, struct net_device *bond_dev)
|
3381 |
|
|
{
|
3382 |
|
|
struct bonding *event_bond = bond_dev->priv;
|
3383 |
|
|
|
3384 |
|
|
switch (event) {
|
3385 |
|
|
case NETDEV_CHANGENAME:
|
3386 |
|
|
return bond_event_changename(event_bond);
|
3387 |
|
|
case NETDEV_UNREGISTER:
|
3388 |
|
|
/*
|
3389 |
|
|
* TODO: remove a bond from the list?
|
3390 |
|
|
*/
|
3391 |
|
|
break;
|
3392 |
|
|
default:
|
3393 |
|
|
break;
|
3394 |
|
|
}
|
3395 |
|
|
|
3396 |
|
|
return NOTIFY_DONE;
|
3397 |
|
|
}
|
3398 |
|
|
|
3399 |
|
|
static int bond_slave_netdev_event(unsigned long event, struct net_device *slave_dev)
|
3400 |
|
|
{
|
3401 |
|
|
struct net_device *bond_dev = slave_dev->master;
|
3402 |
|
|
|
3403 |
|
|
switch (event) {
|
3404 |
|
|
case NETDEV_UNREGISTER:
|
3405 |
|
|
if (bond_dev) {
|
3406 |
|
|
bond_release(bond_dev, slave_dev);
|
3407 |
|
|
}
|
3408 |
|
|
break;
|
3409 |
|
|
case NETDEV_CHANGE:
|
3410 |
|
|
/*
|
3411 |
|
|
* TODO: is this what we get if somebody
|
3412 |
|
|
* sets up a hierarchical bond, then rmmod's
|
3413 |
|
|
* one of the slave bonding devices?
|
3414 |
|
|
*/
|
3415 |
|
|
break;
|
3416 |
|
|
case NETDEV_DOWN:
|
3417 |
|
|
/*
|
3418 |
|
|
* ... Or is it this?
|
3419 |
|
|
*/
|
3420 |
|
|
break;
|
3421 |
|
|
case NETDEV_CHANGEMTU:
|
3422 |
|
|
/*
|
3423 |
|
|
* TODO: Should slaves be allowed to
|
3424 |
|
|
* independently alter their MTU? For
|
3425 |
|
|
* an active-backup bond, slaves need
|
3426 |
|
|
* not be the same type of device, so
|
3427 |
|
|
* MTUs may vary. For other modes,
|
3428 |
|
|
* slaves arguably should have the
|
3429 |
|
|
* same MTUs. To do this, we'd need to
|
3430 |
|
|
* take over the slave's change_mtu
|
3431 |
|
|
* function for the duration of their
|
3432 |
|
|
* servitude.
|
3433 |
|
|
*/
|
3434 |
|
|
break;
|
3435 |
|
|
case NETDEV_CHANGENAME:
|
3436 |
|
|
/*
|
3437 |
|
|
* TODO: handle changing the primary's name
|
3438 |
|
|
*/
|
3439 |
|
|
break;
|
3440 |
|
|
default:
|
3441 |
|
|
break;
|
3442 |
|
|
}
|
3443 |
|
|
|
3444 |
|
|
return NOTIFY_DONE;
|
3445 |
|
|
}
|
3446 |
|
|
|
3447 |
|
|
/*
|
3448 |
|
|
* bond_netdev_event: handle netdev notifier chain events.
|
3449 |
|
|
*
|
3450 |
|
|
* This function receives events for the netdev chain. The caller (an
|
3451 |
|
|
* ioctl handler calling notifier_call_chain) holds the necessary
|
3452 |
|
|
* locks for us to safely manipulate the slave devices (RTNL lock,
|
3453 |
|
|
* dev_probe_lock).
|
3454 |
|
|
*/
|
3455 |
|
|
static int bond_netdev_event(struct notifier_block *this, unsigned long event, void *ptr)
|
3456 |
|
|
{
|
3457 |
|
|
struct net_device *event_dev = (struct net_device *)ptr;
|
3458 |
|
|
|
3459 |
|
|
dprintk("event_dev: %s, event: %lx\n",
|
3460 |
|
|
(event_dev ? event_dev->name : "None"),
|
3461 |
|
|
event);
|
3462 |
|
|
|
3463 |
|
|
if (event_dev->flags & IFF_MASTER) {
|
3464 |
|
|
dprintk("IFF_MASTER\n");
|
3465 |
|
|
return bond_master_netdev_event(event, event_dev);
|
3466 |
|
|
}
|
3467 |
|
|
|
3468 |
|
|
if (event_dev->flags & IFF_SLAVE) {
|
3469 |
|
|
dprintk("IFF_SLAVE\n");
|
3470 |
|
|
return bond_slave_netdev_event(event, event_dev);
|
3471 |
|
|
}
|
3472 |
|
|
|
3473 |
|
|
return NOTIFY_DONE;
|
3474 |
|
|
}
|
3475 |
|
|
|
3476 |
|
|
static struct notifier_block bond_netdev_notifier = {
|
3477 |
|
|
.notifier_call = bond_netdev_event,
|
3478 |
|
|
};
|
3479 |
|
|
|
3480 |
|
|
/*-------------------------- Packet type handling ---------------------------*/
|
3481 |
|
|
|
3482 |
|
|
/* register to receive lacpdus on a bond */
|
3483 |
|
|
static void bond_register_lacpdu(struct bonding *bond)
|
3484 |
|
|
{
|
3485 |
|
|
struct packet_type *pk_type = &(BOND_AD_INFO(bond).ad_pkt_type);
|
3486 |
|
|
|
3487 |
|
|
/* initialize packet type */
|
3488 |
|
|
pk_type->type = PKT_TYPE_LACPDU;
|
3489 |
|
|
pk_type->dev = bond->dev;
|
3490 |
|
|
pk_type->func = bond_3ad_lacpdu_recv;
|
3491 |
|
|
pk_type->data = (void*)1; /* understand shared skbs */
|
3492 |
|
|
|
3493 |
|
|
dev_add_pack(pk_type);
|
3494 |
|
|
}
|
3495 |
|
|
|
3496 |
|
|
/* unregister to receive lacpdus on a bond */
|
3497 |
|
|
static void bond_unregister_lacpdu(struct bonding *bond)
|
3498 |
|
|
{
|
3499 |
|
|
dev_remove_pack(&(BOND_AD_INFO(bond).ad_pkt_type));
|
3500 |
|
|
}
|
3501 |
|
|
|
3502 |
|
|
/*-------------------------- Device entry points ----------------------------*/
|
3503 |
|
|
|
3504 |
|
|
static int bond_open(struct net_device *bond_dev)
|
3505 |
|
|
{
|
3506 |
|
|
struct bonding *bond = bond_dev->priv;
|
3507 |
|
|
struct timer_list *mii_timer = &bond->mii_timer;
|
3508 |
|
|
struct timer_list *arp_timer = &bond->arp_timer;
|
3509 |
|
|
|
3510 |
|
|
bond->kill_timers = 0;
|
3511 |
|
|
|
3512 |
|
|
if ((bond->params.mode == BOND_MODE_TLB) ||
|
3513 |
|
|
(bond->params.mode == BOND_MODE_ALB)) {
|
3514 |
|
|
struct timer_list *alb_timer = &(BOND_ALB_INFO(bond).alb_timer);
|
3515 |
|
|
|
3516 |
|
|
/* bond_alb_initialize must be called before the timer
|
3517 |
|
|
* is started.
|
3518 |
|
|
*/
|
3519 |
|
|
if (bond_alb_initialize(bond, (bond->params.mode == BOND_MODE_ALB))) {
|
3520 |
|
|
/* something went wrong - fail the open operation */
|
3521 |
|
|
return -1;
|
3522 |
|
|
}
|
3523 |
|
|
|
3524 |
|
|
init_timer(alb_timer);
|
3525 |
|
|
alb_timer->expires = jiffies + 1;
|
3526 |
|
|
alb_timer->data = (unsigned long)bond;
|
3527 |
|
|
alb_timer->function = (void *)&bond_alb_monitor;
|
3528 |
|
|
add_timer(alb_timer);
|
3529 |
|
|
}
|
3530 |
|
|
|
3531 |
|
|
if (bond->params.miimon) { /* link check interval, in milliseconds. */
|
3532 |
|
|
init_timer(mii_timer);
|
3533 |
|
|
mii_timer->expires = jiffies + 1;
|
3534 |
|
|
mii_timer->data = (unsigned long)bond_dev;
|
3535 |
|
|
mii_timer->function = (void *)&bond_mii_monitor;
|
3536 |
|
|
add_timer(mii_timer);
|
3537 |
|
|
}
|
3538 |
|
|
|
3539 |
|
|
if (bond->params.arp_interval) { /* arp interval, in milliseconds. */
|
3540 |
|
|
init_timer(arp_timer);
|
3541 |
|
|
arp_timer->expires = jiffies + 1;
|
3542 |
|
|
arp_timer->data = (unsigned long)bond_dev;
|
3543 |
|
|
if (bond->params.mode == BOND_MODE_ACTIVEBACKUP) {
|
3544 |
|
|
arp_timer->function = (void *)&bond_activebackup_arp_mon;
|
3545 |
|
|
} else {
|
3546 |
|
|
arp_timer->function = (void *)&bond_loadbalance_arp_mon;
|
3547 |
|
|
}
|
3548 |
|
|
add_timer(arp_timer);
|
3549 |
|
|
}
|
3550 |
|
|
|
3551 |
|
|
if (bond->params.mode == BOND_MODE_8023AD) {
|
3552 |
|
|
struct timer_list *ad_timer = &(BOND_AD_INFO(bond).ad_timer);
|
3553 |
|
|
init_timer(ad_timer);
|
3554 |
|
|
ad_timer->expires = jiffies + 1;
|
3555 |
|
|
ad_timer->data = (unsigned long)bond;
|
3556 |
|
|
ad_timer->function = (void *)&bond_3ad_state_machine_handler;
|
3557 |
|
|
add_timer(ad_timer);
|
3558 |
|
|
|
3559 |
|
|
/* register to receive LACPDUs */
|
3560 |
|
|
bond_register_lacpdu(bond);
|
3561 |
|
|
}
|
3562 |
|
|
|
3563 |
|
|
return 0;
|
3564 |
|
|
}
|
3565 |
|
|
|
3566 |
|
|
static int bond_close(struct net_device *bond_dev)
|
3567 |
|
|
{
|
3568 |
|
|
struct bonding *bond = bond_dev->priv;
|
3569 |
|
|
|
3570 |
|
|
write_lock_bh(&bond->lock);
|
3571 |
|
|
|
3572 |
|
|
bond_mc_list_destroy(bond);
|
3573 |
|
|
|
3574 |
|
|
if (bond->params.mode == BOND_MODE_8023AD) {
|
3575 |
|
|
/* Unregister the receive of LACPDUs */
|
3576 |
|
|
bond_unregister_lacpdu(bond);
|
3577 |
|
|
}
|
3578 |
|
|
|
3579 |
|
|
/* signal timers not to re-arm */
|
3580 |
|
|
bond->kill_timers = 1;
|
3581 |
|
|
|
3582 |
|
|
write_unlock_bh(&bond->lock);
|
3583 |
|
|
|
3584 |
|
|
/* del_timer_sync must run without holding the bond->lock
|
3585 |
|
|
* because a running timer might be trying to hold it too
|
3586 |
|
|
*/
|
3587 |
|
|
|
3588 |
|
|
if (bond->params.miimon) { /* link check interval, in milliseconds. */
|
3589 |
|
|
del_timer_sync(&bond->mii_timer);
|
3590 |
|
|
}
|
3591 |
|
|
|
3592 |
|
|
if (bond->params.arp_interval) { /* arp interval, in milliseconds. */
|
3593 |
|
|
del_timer_sync(&bond->arp_timer);
|
3594 |
|
|
}
|
3595 |
|
|
|
3596 |
|
|
switch (bond->params.mode) {
|
3597 |
|
|
case BOND_MODE_8023AD:
|
3598 |
|
|
del_timer_sync(&(BOND_AD_INFO(bond).ad_timer));
|
3599 |
|
|
break;
|
3600 |
|
|
case BOND_MODE_TLB:
|
3601 |
|
|
case BOND_MODE_ALB:
|
3602 |
|
|
del_timer_sync(&(BOND_ALB_INFO(bond).alb_timer));
|
3603 |
|
|
break;
|
3604 |
|
|
default:
|
3605 |
|
|
break;
|
3606 |
|
|
}
|
3607 |
|
|
|
3608 |
|
|
/* Release the bonded slaves */
|
3609 |
|
|
bond_release_all(bond_dev);
|
3610 |
|
|
|
3611 |
|
|
if ((bond->params.mode == BOND_MODE_TLB) ||
|
3612 |
|
|
(bond->params.mode == BOND_MODE_ALB)) {
|
3613 |
|
|
/* Must be called only after all
|
3614 |
|
|
* slaves have been released
|
3615 |
|
|
*/
|
3616 |
|
|
bond_alb_deinitialize(bond);
|
3617 |
|
|
}
|
3618 |
|
|
|
3619 |
|
|
return 0;
|
3620 |
|
|
}
|
3621 |
|
|
|
3622 |
|
|
static struct net_device_stats *bond_get_stats(struct net_device *bond_dev)
|
3623 |
|
|
{
|
3624 |
|
|
struct bonding *bond = bond_dev->priv;
|
3625 |
|
|
struct net_device_stats *stats = &(bond->stats), *sstats;
|
3626 |
|
|
struct slave *slave;
|
3627 |
|
|
int i;
|
3628 |
|
|
|
3629 |
|
|
memset(stats, 0, sizeof(struct net_device_stats));
|
3630 |
|
|
|
3631 |
|
|
read_lock_bh(&bond->lock);
|
3632 |
|
|
|
3633 |
|
|
bond_for_each_slave(bond, slave, i) {
|
3634 |
|
|
sstats = slave->dev->get_stats(slave->dev);
|
3635 |
|
|
|
3636 |
|
|
stats->rx_packets += sstats->rx_packets;
|
3637 |
|
|
stats->rx_bytes += sstats->rx_bytes;
|
3638 |
|
|
stats->rx_errors += sstats->rx_errors;
|
3639 |
|
|
stats->rx_dropped += sstats->rx_dropped;
|
3640 |
|
|
|
3641 |
|
|
stats->tx_packets += sstats->tx_packets;
|
3642 |
|
|
stats->tx_bytes += sstats->tx_bytes;
|
3643 |
|
|
stats->tx_errors += sstats->tx_errors;
|
3644 |
|
|
stats->tx_dropped += sstats->tx_dropped;
|
3645 |
|
|
|
3646 |
|
|
stats->multicast += sstats->multicast;
|
3647 |
|
|
stats->collisions += sstats->collisions;
|
3648 |
|
|
|
3649 |
|
|
stats->rx_length_errors += sstats->rx_length_errors;
|
3650 |
|
|
stats->rx_over_errors += sstats->rx_over_errors;
|
3651 |
|
|
stats->rx_crc_errors += sstats->rx_crc_errors;
|
3652 |
|
|
stats->rx_frame_errors += sstats->rx_frame_errors;
|
3653 |
|
|
stats->rx_fifo_errors += sstats->rx_fifo_errors;
|
3654 |
|
|
stats->rx_missed_errors += sstats->rx_missed_errors;
|
3655 |
|
|
|
3656 |
|
|
stats->tx_aborted_errors += sstats->tx_aborted_errors;
|
3657 |
|
|
stats->tx_carrier_errors += sstats->tx_carrier_errors;
|
3658 |
|
|
stats->tx_fifo_errors += sstats->tx_fifo_errors;
|
3659 |
|
|
stats->tx_heartbeat_errors += sstats->tx_heartbeat_errors;
|
3660 |
|
|
stats->tx_window_errors += sstats->tx_window_errors;
|
3661 |
|
|
}
|
3662 |
|
|
|
3663 |
|
|
read_unlock_bh(&bond->lock);
|
3664 |
|
|
|
3665 |
|
|
return stats;
|
3666 |
|
|
}
|
3667 |
|
|
|
3668 |
|
|
static int bond_do_ioctl(struct net_device *bond_dev, struct ifreq *ifr, int cmd)
|
3669 |
|
|
{
|
3670 |
|
|
struct net_device *slave_dev = NULL;
|
3671 |
|
|
struct ifbond *u_binfo = NULL, k_binfo;
|
3672 |
|
|
struct ifslave *u_sinfo = NULL, k_sinfo;
|
3673 |
|
|
struct mii_ioctl_data *mii = NULL;
|
3674 |
|
|
int prev_abi_ver = orig_app_abi_ver;
|
3675 |
|
|
int res = 0;
|
3676 |
|
|
|
3677 |
|
|
dprintk("bond_ioctl: master=%s, cmd=%d\n",
|
3678 |
|
|
bond_dev->name, cmd);
|
3679 |
|
|
|
3680 |
|
|
switch (cmd) {
|
3681 |
|
|
case SIOCETHTOOL:
|
3682 |
|
|
return bond_ethtool_ioctl(bond_dev, ifr);
|
3683 |
|
|
case SIOCGMIIPHY:
|
3684 |
|
|
mii = (struct mii_ioctl_data *)&ifr->ifr_data;
|
3685 |
|
|
if (!mii) {
|
3686 |
|
|
return -EINVAL;
|
3687 |
|
|
}
|
3688 |
|
|
mii->phy_id = 0;
|
3689 |
|
|
/* Fall Through */
|
3690 |
|
|
case SIOCGMIIREG:
|
3691 |
|
|
/*
|
3692 |
|
|
* We do this again just in case we were called by SIOCGMIIREG
|
3693 |
|
|
* instead of SIOCGMIIPHY.
|
3694 |
|
|
*/
|
3695 |
|
|
mii = (struct mii_ioctl_data *)&ifr->ifr_data;
|
3696 |
|
|
if (!mii) {
|
3697 |
|
|
return -EINVAL;
|
3698 |
|
|
}
|
3699 |
|
|
|
3700 |
|
|
if (mii->reg_num == 1) {
|
3701 |
|
|
struct bonding *bond = bond_dev->priv;
|
3702 |
|
|
mii->val_out = 0;
|
3703 |
|
|
read_lock_bh(&bond->lock);
|
3704 |
|
|
read_lock(&bond->curr_slave_lock);
|
3705 |
|
|
if (bond->curr_active_slave) {
|
3706 |
|
|
mii->val_out = BMSR_LSTATUS;
|
3707 |
|
|
}
|
3708 |
|
|
read_unlock(&bond->curr_slave_lock);
|
3709 |
|
|
read_unlock_bh(&bond->lock);
|
3710 |
|
|
}
|
3711 |
|
|
|
3712 |
|
|
return 0;
|
3713 |
|
|
case BOND_INFO_QUERY_OLD:
|
3714 |
|
|
case SIOCBONDINFOQUERY:
|
3715 |
|
|
u_binfo = (struct ifbond *)ifr->ifr_data;
|
3716 |
|
|
|
3717 |
|
|
if (copy_from_user(&k_binfo, u_binfo, sizeof(ifbond))) {
|
3718 |
|
|
return -EFAULT;
|
3719 |
|
|
}
|
3720 |
|
|
|
3721 |
|
|
res = bond_info_query(bond_dev, &k_binfo);
|
3722 |
|
|
if (res == 0) {
|
3723 |
|
|
if (copy_to_user(u_binfo, &k_binfo, sizeof(ifbond))) {
|
3724 |
|
|
return -EFAULT;
|
3725 |
|
|
}
|
3726 |
|
|
}
|
3727 |
|
|
|
3728 |
|
|
return res;
|
3729 |
|
|
case BOND_SLAVE_INFO_QUERY_OLD:
|
3730 |
|
|
case SIOCBONDSLAVEINFOQUERY:
|
3731 |
|
|
u_sinfo = (struct ifslave *)ifr->ifr_data;
|
3732 |
|
|
|
3733 |
|
|
if (copy_from_user(&k_sinfo, u_sinfo, sizeof(ifslave))) {
|
3734 |
|
|
return -EFAULT;
|
3735 |
|
|
}
|
3736 |
|
|
|
3737 |
|
|
res = bond_slave_info_query(bond_dev, &k_sinfo);
|
3738 |
|
|
if (res == 0) {
|
3739 |
|
|
if (copy_to_user(u_sinfo, &k_sinfo, sizeof(ifslave))) {
|
3740 |
|
|
return -EFAULT;
|
3741 |
|
|
}
|
3742 |
|
|
}
|
3743 |
|
|
|
3744 |
|
|
return res;
|
3745 |
|
|
default:
|
3746 |
|
|
/* Go on */
|
3747 |
|
|
break;
|
3748 |
|
|
}
|
3749 |
|
|
|
3750 |
|
|
if (!capable(CAP_NET_ADMIN)) {
|
3751 |
|
|
return -EPERM;
|
3752 |
|
|
}
|
3753 |
|
|
|
3754 |
|
|
if (orig_app_abi_ver == -1) {
|
3755 |
|
|
/* no orig_app_abi_ver was provided yet, so we'll use the
|
3756 |
|
|
* current one from now on, even if it's 0
|
3757 |
|
|
*/
|
3758 |
|
|
orig_app_abi_ver = app_abi_ver;
|
3759 |
|
|
|
3760 |
|
|
} else if (orig_app_abi_ver != app_abi_ver) {
|
3761 |
|
|
printk(KERN_ERR DRV_NAME
|
3762 |
|
|
": Error: already using ifenslave ABI version %d; to "
|
3763 |
|
|
"upgrade ifenslave to version %d, you must first "
|
3764 |
|
|
"reload bonding.\n",
|
3765 |
|
|
orig_app_abi_ver, app_abi_ver);
|
3766 |
|
|
return -EINVAL;
|
3767 |
|
|
}
|
3768 |
|
|
|
3769 |
|
|
slave_dev = dev_get_by_name(ifr->ifr_slave);
|
3770 |
|
|
|
3771 |
|
|
dprintk("slave_dev=%p: \n", slave_dev);
|
3772 |
|
|
|
3773 |
|
|
if (!slave_dev) {
|
3774 |
|
|
res = -ENODEV;
|
3775 |
|
|
} else {
|
3776 |
|
|
dprintk("slave_dev->name=%s: \n", slave_dev->name);
|
3777 |
|
|
switch (cmd) {
|
3778 |
|
|
case BOND_ENSLAVE_OLD:
|
3779 |
|
|
case SIOCBONDENSLAVE:
|
3780 |
|
|
res = bond_enslave(bond_dev, slave_dev);
|
3781 |
|
|
break;
|
3782 |
|
|
case BOND_RELEASE_OLD:
|
3783 |
|
|
case SIOCBONDRELEASE:
|
3784 |
|
|
res = bond_release(bond_dev, slave_dev);
|
3785 |
|
|
break;
|
3786 |
|
|
case BOND_SETHWADDR_OLD:
|
3787 |
|
|
case SIOCBONDSETHWADDR:
|
3788 |
|
|
res = bond_sethwaddr(bond_dev, slave_dev);
|
3789 |
|
|
break;
|
3790 |
|
|
case BOND_CHANGE_ACTIVE_OLD:
|
3791 |
|
|
case SIOCBONDCHANGEACTIVE:
|
3792 |
|
|
res = bond_ioctl_change_active(bond_dev, slave_dev);
|
3793 |
|
|
break;
|
3794 |
|
|
default:
|
3795 |
|
|
res = -EOPNOTSUPP;
|
3796 |
|
|
}
|
3797 |
|
|
|
3798 |
|
|
dev_put(slave_dev);
|
3799 |
|
|
}
|
3800 |
|
|
|
3801 |
|
|
if (res < 0) {
|
3802 |
|
|
/* The ioctl failed, so there's no point in changing the
|
3803 |
|
|
* orig_app_abi_ver. We'll restore it's value just in case
|
3804 |
|
|
* we've changed it earlier in this function.
|
3805 |
|
|
*/
|
3806 |
|
|
orig_app_abi_ver = prev_abi_ver;
|
3807 |
|
|
}
|
3808 |
|
|
|
3809 |
|
|
return res;
|
3810 |
|
|
}
|
3811 |
|
|
|
3812 |
|
|
static void bond_set_multicast_list(struct net_device *bond_dev)
|
3813 |
|
|
{
|
3814 |
|
|
struct bonding *bond = bond_dev->priv;
|
3815 |
|
|
struct dev_mc_list *dmi;
|
3816 |
|
|
|
3817 |
|
|
write_lock_bh(&bond->lock);
|
3818 |
|
|
|
3819 |
|
|
/*
|
3820 |
|
|
* Do promisc before checking multicast_mode
|
3821 |
|
|
*/
|
3822 |
|
|
if ((bond_dev->flags & IFF_PROMISC) && !(bond->flags & IFF_PROMISC)) {
|
3823 |
|
|
bond_set_promiscuity(bond, 1);
|
3824 |
|
|
}
|
3825 |
|
|
|
3826 |
|
|
if (!(bond_dev->flags & IFF_PROMISC) && (bond->flags & IFF_PROMISC)) {
|
3827 |
|
|
bond_set_promiscuity(bond, -1);
|
3828 |
|
|
}
|
3829 |
|
|
|
3830 |
|
|
/* set allmulti flag to slaves */
|
3831 |
|
|
if ((bond_dev->flags & IFF_ALLMULTI) && !(bond->flags & IFF_ALLMULTI)) {
|
3832 |
|
|
bond_set_allmulti(bond, 1);
|
3833 |
|
|
}
|
3834 |
|
|
|
3835 |
|
|
if (!(bond_dev->flags & IFF_ALLMULTI) && (bond->flags & IFF_ALLMULTI)) {
|
3836 |
|
|
bond_set_allmulti(bond, -1);
|
3837 |
|
|
}
|
3838 |
|
|
|
3839 |
|
|
bond->flags = bond_dev->flags;
|
3840 |
|
|
|
3841 |
|
|
/* looking for addresses to add to slaves' mc list */
|
3842 |
|
|
for (dmi = bond_dev->mc_list; dmi; dmi = dmi->next) {
|
3843 |
|
|
if (!bond_mc_list_find_dmi(dmi, bond->mc_list)) {
|
3844 |
|
|
bond_mc_add(bond, dmi->dmi_addr, dmi->dmi_addrlen);
|
3845 |
|
|
}
|
3846 |
|
|
}
|
3847 |
|
|
|
3848 |
|
|
/* looking for addresses to delete from slaves' list */
|
3849 |
|
|
for (dmi = bond->mc_list; dmi; dmi = dmi->next) {
|
3850 |
|
|
if (!bond_mc_list_find_dmi(dmi, bond_dev->mc_list)) {
|
3851 |
|
|
bond_mc_delete(bond, dmi->dmi_addr, dmi->dmi_addrlen);
|
3852 |
|
|
}
|
3853 |
|
|
}
|
3854 |
|
|
|
3855 |
|
|
/* save master's multicast list */
|
3856 |
|
|
bond_mc_list_destroy(bond);
|
3857 |
|
|
bond_mc_list_copy(bond_dev->mc_list, bond, GFP_ATOMIC);
|
3858 |
|
|
|
3859 |
|
|
write_unlock_bh(&bond->lock);
|
3860 |
|
|
}
|
3861 |
|
|
|
3862 |
|
|
/*
|
3863 |
|
|
* Change the MTU of all of a master's slaves to match the master
|
3864 |
|
|
*/
|
3865 |
|
|
static int bond_change_mtu(struct net_device *bond_dev, int new_mtu)
|
3866 |
|
|
{
|
3867 |
|
|
struct bonding *bond = bond_dev->priv;
|
3868 |
|
|
struct slave *slave, *stop_at;
|
3869 |
|
|
int res = 0;
|
3870 |
|
|
int i;
|
3871 |
|
|
|
3872 |
|
|
dprintk("bond=%p, name=%s, new_mtu=%d\n", bond,
|
3873 |
|
|
(bond_dev ? bond_dev->name : "None"), new_mtu);
|
3874 |
|
|
|
3875 |
|
|
/* Can't hold bond->lock with bh disabled here since
|
3876 |
|
|
* some base drivers panic. On the other hand we can't
|
3877 |
|
|
* hold bond->lock without bh disabled because we'll
|
3878 |
|
|
* deadlock. The only solution is to rely on the fact
|
3879 |
|
|
* that we're under rtnl_lock here, and the slaves
|
3880 |
|
|
* list won't change. This doesn't solve the problem
|
3881 |
|
|
* of setting the slave's MTU while it is
|
3882 |
|
|
* transmitting, but the assumption is that the base
|
3883 |
|
|
* driver can handle that.
|
3884 |
|
|
*
|
3885 |
|
|
* TODO: figure out a way to safely iterate the slaves
|
3886 |
|
|
* list, but without holding a lock around the actual
|
3887 |
|
|
* call to the base driver.
|
3888 |
|
|
*/
|
3889 |
|
|
|
3890 |
|
|
bond_for_each_slave(bond, slave, i) {
|
3891 |
|
|
dprintk("s %p s->p %p c_m %p\n", slave,
|
3892 |
|
|
slave->prev, slave->dev->change_mtu);
|
3893 |
|
|
if (slave->dev->change_mtu) {
|
3894 |
|
|
res = slave->dev->change_mtu(slave->dev, new_mtu);
|
3895 |
|
|
} else {
|
3896 |
|
|
slave->dev->mtu = new_mtu;
|
3897 |
|
|
res = 0;
|
3898 |
|
|
}
|
3899 |
|
|
|
3900 |
|
|
if (res) {
|
3901 |
|
|
/* If we failed to set the slave's mtu to the new value
|
3902 |
|
|
* we must abort the operation even in ACTIVE_BACKUP
|
3903 |
|
|
* mode, because if we allow the backup slaves to have
|
3904 |
|
|
* different mtu values than the active slave we'll
|
3905 |
|
|
* need to change their mtu when doing a failover. That
|
3906 |
|
|
* means changing their mtu from timer context, which
|
3907 |
|
|
* is probably not a good idea.
|
3908 |
|
|
*/
|
3909 |
|
|
dprintk("err %d %s\n", res, slave->dev->name);
|
3910 |
|
|
goto unwind;
|
3911 |
|
|
}
|
3912 |
|
|
}
|
3913 |
|
|
|
3914 |
|
|
bond_dev->mtu = new_mtu;
|
3915 |
|
|
|
3916 |
|
|
return 0;
|
3917 |
|
|
|
3918 |
|
|
unwind:
|
3919 |
|
|
/* unwind from head to the slave that failed */
|
3920 |
|
|
stop_at = slave;
|
3921 |
|
|
bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) {
|
3922 |
|
|
int tmp_res;
|
3923 |
|
|
|
3924 |
|
|
if (slave->dev->change_mtu) {
|
3925 |
|
|
tmp_res = slave->dev->change_mtu(slave->dev, bond_dev->mtu);
|
3926 |
|
|
if (tmp_res) {
|
3927 |
|
|
dprintk("unwind err %d dev %s\n", tmp_res,
|
3928 |
|
|
slave->dev->name);
|
3929 |
|
|
}
|
3930 |
|
|
} else {
|
3931 |
|
|
slave->dev->mtu = bond_dev->mtu;
|
3932 |
|
|
}
|
3933 |
|
|
}
|
3934 |
|
|
|
3935 |
|
|
return res;
|
3936 |
|
|
}
|
3937 |
|
|
|
3938 |
|
|
/*
|
3939 |
|
|
* Change HW address
|
3940 |
|
|
*
|
3941 |
|
|
* Note that many devices must be down to change the HW address, and
|
3942 |
|
|
* downing the master releases all slaves. We can make bonds full of
|
3943 |
|
|
* bonding devices to test this, however.
|
3944 |
|
|
*/
|
3945 |
|
|
static int bond_set_mac_address(struct net_device *bond_dev, void *addr)
|
3946 |
|
|
{
|
3947 |
|
|
struct bonding *bond = bond_dev->priv;
|
3948 |
|
|
struct sockaddr *sa = addr, tmp_sa;
|
3949 |
|
|
struct slave *slave, *stop_at;
|
3950 |
|
|
int res = 0;
|
3951 |
|
|
int i;
|
3952 |
|
|
|
3953 |
|
|
dprintk("bond=%p, name=%s\n", bond, (bond_dev ? bond_dev->name : "None"));
|
3954 |
|
|
|
3955 |
|
|
if (!is_valid_ether_addr(sa->sa_data)) {
|
3956 |
|
|
return -EADDRNOTAVAIL;
|
3957 |
|
|
}
|
3958 |
|
|
|
3959 |
|
|
/* Can't hold bond->lock with bh disabled here since
|
3960 |
|
|
* some base drivers panic. On the other hand we can't
|
3961 |
|
|
* hold bond->lock without bh disabled because we'll
|
3962 |
|
|
* deadlock. The only solution is to rely on the fact
|
3963 |
|
|
* that we're under rtnl_lock here, and the slaves
|
3964 |
|
|
* list won't change. This doesn't solve the problem
|
3965 |
|
|
* of setting the slave's hw address while it is
|
3966 |
|
|
* transmitting, but the assumption is that the base
|
3967 |
|
|
* driver can handle that.
|
3968 |
|
|
*
|
3969 |
|
|
* TODO: figure out a way to safely iterate the slaves
|
3970 |
|
|
* list, but without holding a lock around the actual
|
3971 |
|
|
* call to the base driver.
|
3972 |
|
|
*/
|
3973 |
|
|
|
3974 |
|
|
bond_for_each_slave(bond, slave, i) {
|
3975 |
|
|
dprintk("slave %p %s\n", slave, slave->dev->name);
|
3976 |
|
|
|
3977 |
|
|
if (slave->dev->set_mac_address == NULL) {
|
3978 |
|
|
res = -EOPNOTSUPP;
|
3979 |
|
|
dprintk("EOPNOTSUPP %s\n", slave->dev->name);
|
3980 |
|
|
goto unwind;
|
3981 |
|
|
}
|
3982 |
|
|
|
3983 |
|
|
res = slave->dev->set_mac_address(slave->dev, addr);
|
3984 |
|
|
if (res) {
|
3985 |
|
|
/* TODO: consider downing the slave
|
3986 |
|
|
* and retry ?
|
3987 |
|
|
* User should expect communications
|
3988 |
|
|
* breakage anyway until ARP finish
|
3989 |
|
|
* updating, so...
|
3990 |
|
|
*/
|
3991 |
|
|
dprintk("err %d %s\n", res, slave->dev->name);
|
3992 |
|
|
goto unwind;
|
3993 |
|
|
}
|
3994 |
|
|
}
|
3995 |
|
|
|
3996 |
|
|
/* success */
|
3997 |
|
|
memcpy(bond_dev->dev_addr, sa->sa_data, bond_dev->addr_len);
|
3998 |
|
|
return 0;
|
3999 |
|
|
|
4000 |
|
|
unwind:
|
4001 |
|
|
memcpy(tmp_sa.sa_data, bond_dev->dev_addr, bond_dev->addr_len);
|
4002 |
|
|
tmp_sa.sa_family = bond_dev->type;
|
4003 |
|
|
|
4004 |
|
|
/* unwind from head to the slave that failed */
|
4005 |
|
|
stop_at = slave;
|
4006 |
|
|
bond_for_each_slave_from_to(bond, slave, i, bond->first_slave, stop_at) {
|
4007 |
|
|
int tmp_res;
|
4008 |
|
|
|
4009 |
|
|
tmp_res = slave->dev->set_mac_address(slave->dev, &tmp_sa);
|
4010 |
|
|
if (tmp_res) {
|
4011 |
|
|
dprintk("unwind err %d dev %s\n", tmp_res,
|
4012 |
|
|
slave->dev->name);
|
4013 |
|
|
}
|
4014 |
|
|
}
|
4015 |
|
|
|
4016 |
|
|
return res;
|
4017 |
|
|
}
|
4018 |
|
|
|
4019 |
|
|
static int bond_xmit_roundrobin(struct sk_buff *skb, struct net_device *bond_dev)
|
4020 |
|
|
{
|
4021 |
|
|
struct bonding *bond = bond_dev->priv;
|
4022 |
|
|
struct slave *slave, *start_at;
|
4023 |
|
|
int i;
|
4024 |
|
|
int res = 1;
|
4025 |
|
|
|
4026 |
|
|
read_lock(&bond->lock);
|
4027 |
|
|
|
4028 |
|
|
if (!BOND_IS_OK(bond)) {
|
4029 |
|
|
goto out;
|
4030 |
|
|
}
|
4031 |
|
|
|
4032 |
|
|
read_lock(&bond->curr_slave_lock);
|
4033 |
|
|
slave = start_at = bond->curr_active_slave;
|
4034 |
|
|
read_unlock(&bond->curr_slave_lock);
|
4035 |
|
|
|
4036 |
|
|
if (!slave) {
|
4037 |
|
|
goto out;
|
4038 |
|
|
}
|
4039 |
|
|
|
4040 |
|
|
bond_for_each_slave_from(bond, slave, i, start_at) {
|
4041 |
|
|
if (IS_UP(slave->dev) &&
|
4042 |
|
|
(slave->link == BOND_LINK_UP) &&
|
4043 |
|
|
(slave->state == BOND_STATE_ACTIVE)) {
|
4044 |
|
|
res = bond_dev_queue_xmit(bond, skb, slave->dev);
|
4045 |
|
|
|
4046 |
|
|
write_lock(&bond->curr_slave_lock);
|
4047 |
|
|
bond->curr_active_slave = slave->next;
|
4048 |
|
|
write_unlock(&bond->curr_slave_lock);
|
4049 |
|
|
|
4050 |
|
|
break;
|
4051 |
|
|
}
|
4052 |
|
|
}
|
4053 |
|
|
|
4054 |
|
|
|
4055 |
|
|
out:
|
4056 |
|
|
if (res) {
|
4057 |
|
|
/* no suitable interface, frame not sent */
|
4058 |
|
|
dev_kfree_skb(skb);
|
4059 |
|
|
}
|
4060 |
|
|
read_unlock(&bond->lock);
|
4061 |
|
|
return 0;
|
4062 |
|
|
}
|
4063 |
|
|
|
4064 |
|
|
/*
|
4065 |
|
|
* in active-backup mode, we know that bond->curr_active_slave is always valid if
|
4066 |
|
|
* the bond has a usable interface.
|
4067 |
|
|
*/
|
4068 |
|
|
static int bond_xmit_activebackup(struct sk_buff *skb, struct net_device *bond_dev)
|
4069 |
|
|
{
|
4070 |
|
|
struct bonding *bond = bond_dev->priv;
|
4071 |
|
|
int res = 1;
|
4072 |
|
|
|
4073 |
|
|
/* if we are sending arp packets, try to at least
|
4074 |
|
|
identify our own ip address */
|
4075 |
|
|
if (bond->params.arp_interval && !my_ip &&
|
4076 |
|
|
(skb->protocol == __constant_htons(ETH_P_ARP))) {
|
4077 |
|
|
char *the_ip = (char *)skb->data +
|
4078 |
|
|
sizeof(struct ethhdr) +
|
4079 |
|
|
sizeof(struct arphdr) +
|
4080 |
|
|
ETH_ALEN;
|
4081 |
|
|
memcpy(&my_ip, the_ip, 4);
|
4082 |
|
|
}
|
4083 |
|
|
|
4084 |
|
|
read_lock(&bond->lock);
|
4085 |
|
|
read_lock(&bond->curr_slave_lock);
|
4086 |
|
|
|
4087 |
|
|
if (!BOND_IS_OK(bond)) {
|
4088 |
|
|
goto out;
|
4089 |
|
|
}
|
4090 |
|
|
|
4091 |
|
|
if (bond->curr_active_slave) { /* one usable interface */
|
4092 |
|
|
res = bond_dev_queue_xmit(bond, skb, bond->curr_active_slave->dev);
|
4093 |
|
|
}
|
4094 |
|
|
|
4095 |
|
|
out:
|
4096 |
|
|
if (res) {
|
4097 |
|
|
/* no suitable interface, frame not sent */
|
4098 |
|
|
dev_kfree_skb(skb);
|
4099 |
|
|
}
|
4100 |
|
|
read_unlock(&bond->curr_slave_lock);
|
4101 |
|
|
read_unlock(&bond->lock);
|
4102 |
|
|
return 0;
|
4103 |
|
|
}
|
4104 |
|
|
|
4105 |
|
|
/*
|
4106 |
|
|
* in XOR mode, we determine the output device by performing xor on
|
4107 |
|
|
* the source and destination hw adresses. If this device is not
|
4108 |
|
|
* enabled, find the next slave following this xor slave.
|
4109 |
|
|
*/
|
4110 |
|
|
static int bond_xmit_xor(struct sk_buff *skb, struct net_device *bond_dev)
|
4111 |
|
|
{
|
4112 |
|
|
struct bonding *bond = bond_dev->priv;
|
4113 |
|
|
struct ethhdr *data = (struct ethhdr *)skb->data;
|
4114 |
|
|
struct slave *slave, *start_at;
|
4115 |
|
|
int slave_no;
|
4116 |
|
|
int i;
|
4117 |
|
|
int res = 1;
|
4118 |
|
|
|
4119 |
|
|
read_lock(&bond->lock);
|
4120 |
|
|
|
4121 |
|
|
if (!BOND_IS_OK(bond)) {
|
4122 |
|
|
goto out;
|
4123 |
|
|
}
|
4124 |
|
|
|
4125 |
|
|
slave_no = (data->h_dest[5]^bond_dev->dev_addr[5]) % bond->slave_cnt;
|
4126 |
|
|
|
4127 |
|
|
bond_for_each_slave(bond, slave, i) {
|
4128 |
|
|
slave_no--;
|
4129 |
|
|
if (slave_no < 0) {
|
4130 |
|
|
break;
|
4131 |
|
|
}
|
4132 |
|
|
}
|
4133 |
|
|
|
4134 |
|
|
start_at = slave;
|
4135 |
|
|
|
4136 |
|
|
bond_for_each_slave_from(bond, slave, i, start_at) {
|
4137 |
|
|
if (IS_UP(slave->dev) &&
|
4138 |
|
|
(slave->link == BOND_LINK_UP) &&
|
4139 |
|
|
(slave->state == BOND_STATE_ACTIVE)) {
|
4140 |
|
|
res = bond_dev_queue_xmit(bond, skb, slave->dev);
|
4141 |
|
|
break;
|
4142 |
|
|
}
|
4143 |
|
|
}
|
4144 |
|
|
|
4145 |
|
|
out:
|
4146 |
|
|
if (res) {
|
4147 |
|
|
/* no suitable interface, frame not sent */
|
4148 |
|
|
dev_kfree_skb(skb);
|
4149 |
|
|
}
|
4150 |
|
|
read_unlock(&bond->lock);
|
4151 |
|
|
return 0;
|
4152 |
|
|
}
|
4153 |
|
|
|
4154 |
|
|
/*
|
4155 |
|
|
* in broadcast mode, we send everything to all usable interfaces.
|
4156 |
|
|
*/
|
4157 |
|
|
static int bond_xmit_broadcast(struct sk_buff *skb, struct net_device *bond_dev)
|
4158 |
|
|
{
|
4159 |
|
|
struct bonding *bond = bond_dev->priv;
|
4160 |
|
|
struct slave *slave, *start_at;
|
4161 |
|
|
struct net_device *tx_dev = NULL;
|
4162 |
|
|
int i;
|
4163 |
|
|
int res = 1;
|
4164 |
|
|
|
4165 |
|
|
read_lock(&bond->lock);
|
4166 |
|
|
|
4167 |
|
|
if (!BOND_IS_OK(bond)) {
|
4168 |
|
|
goto out;
|
4169 |
|
|
}
|
4170 |
|
|
|
4171 |
|
|
read_lock(&bond->curr_slave_lock);
|
4172 |
|
|
start_at = bond->curr_active_slave;
|
4173 |
|
|
read_unlock(&bond->curr_slave_lock);
|
4174 |
|
|
|
4175 |
|
|
if (!start_at) {
|
4176 |
|
|
goto out;
|
4177 |
|
|
}
|
4178 |
|
|
|
4179 |
|
|
bond_for_each_slave_from(bond, slave, i, start_at) {
|
4180 |
|
|
if (IS_UP(slave->dev) &&
|
4181 |
|
|
(slave->link == BOND_LINK_UP) &&
|
4182 |
|
|
(slave->state == BOND_STATE_ACTIVE)) {
|
4183 |
|
|
if (tx_dev) {
|
4184 |
|
|
struct sk_buff *skb2 = skb_clone(skb, GFP_ATOMIC);
|
4185 |
|
|
if (!skb2) {
|
4186 |
|
|
printk(KERN_ERR DRV_NAME
|
4187 |
|
|
": Error: bond_xmit_broadcast(): "
|
4188 |
|
|
"skb_clone() failed\n");
|
4189 |
|
|
continue;
|
4190 |
|
|
}
|
4191 |
|
|
|
4192 |
|
|
res = bond_dev_queue_xmit(bond, skb2, tx_dev);
|
4193 |
|
|
if (res) {
|
4194 |
|
|
dev_kfree_skb(skb2);
|
4195 |
|
|
continue;
|
4196 |
|
|
}
|
4197 |
|
|
}
|
4198 |
|
|
tx_dev = slave->dev;
|
4199 |
|
|
}
|
4200 |
|
|
}
|
4201 |
|
|
|
4202 |
|
|
if (tx_dev) {
|
4203 |
|
|
res = bond_dev_queue_xmit(bond, skb, tx_dev);
|
4204 |
|
|
}
|
4205 |
|
|
|
4206 |
|
|
out:
|
4207 |
|
|
if (res) {
|
4208 |
|
|
/* no suitable interface, frame not sent */
|
4209 |
|
|
dev_kfree_skb(skb);
|
4210 |
|
|
}
|
4211 |
|
|
/* frame sent to all suitable interfaces */
|
4212 |
|
|
read_unlock(&bond->lock);
|
4213 |
|
|
return 0;
|
4214 |
|
|
}
|
4215 |
|
|
|
4216 |
|
|
#ifdef CONFIG_NET_FASTROUTE
|
4217 |
|
|
static int bond_accept_fastpath(struct net_device *bond_dev, struct dst_entry *dst)
|
4218 |
|
|
{
|
4219 |
|
|
return -1;
|
4220 |
|
|
}
|
4221 |
|
|
#endif
|
4222 |
|
|
|
4223 |
|
|
/*------------------------- Device initialization ---------------------------*/
|
4224 |
|
|
|
4225 |
|
|
/*
|
4226 |
|
|
* set bond mode specific net device operations
|
4227 |
|
|
*/
|
4228 |
|
|
static inline void bond_set_mode_ops(struct net_device *bond_dev, int mode)
|
4229 |
|
|
{
|
4230 |
|
|
switch (mode) {
|
4231 |
|
|
case BOND_MODE_ROUNDROBIN:
|
4232 |
|
|
bond_dev->hard_start_xmit = bond_xmit_roundrobin;
|
4233 |
|
|
break;
|
4234 |
|
|
case BOND_MODE_ACTIVEBACKUP:
|
4235 |
|
|
bond_dev->hard_start_xmit = bond_xmit_activebackup;
|
4236 |
|
|
break;
|
4237 |
|
|
case BOND_MODE_XOR:
|
4238 |
|
|
bond_dev->hard_start_xmit = bond_xmit_xor;
|
4239 |
|
|
break;
|
4240 |
|
|
case BOND_MODE_BROADCAST:
|
4241 |
|
|
bond_dev->hard_start_xmit = bond_xmit_broadcast;
|
4242 |
|
|
break;
|
4243 |
|
|
case BOND_MODE_8023AD:
|
4244 |
|
|
bond_dev->hard_start_xmit = bond_3ad_xmit_xor;
|
4245 |
|
|
break;
|
4246 |
|
|
case BOND_MODE_TLB:
|
4247 |
|
|
case BOND_MODE_ALB:
|
4248 |
|
|
bond_dev->hard_start_xmit = bond_alb_xmit;
|
4249 |
|
|
bond_dev->set_mac_address = bond_alb_set_mac_address;
|
4250 |
|
|
break;
|
4251 |
|
|
default:
|
4252 |
|
|
/* Should never happen, mode already checked */
|
4253 |
|
|
printk(KERN_ERR DRV_NAME
|
4254 |
|
|
": Error: Unknown bonding mode %d\n",
|
4255 |
|
|
mode);
|
4256 |
|
|
break;
|
4257 |
|
|
}
|
4258 |
|
|
}
|
4259 |
|
|
|
4260 |
|
|
/*
|
4261 |
|
|
* Does not allocate but creates a /proc entry.
|
4262 |
|
|
* Allowed to fail.
|
4263 |
|
|
*/
|
4264 |
|
|
static int __init bond_init(struct net_device *bond_dev, struct bond_params *params)
|
4265 |
|
|
{
|
4266 |
|
|
struct bonding *bond = bond_dev->priv;
|
4267 |
|
|
|
4268 |
|
|
dprintk("Begin bond_init for %s\n", bond_dev->name);
|
4269 |
|
|
|
4270 |
|
|
/* initialize rwlocks */
|
4271 |
|
|
rwlock_init(&bond->lock);
|
4272 |
|
|
rwlock_init(&bond->curr_slave_lock);
|
4273 |
|
|
|
4274 |
|
|
bond->params = *params; /* copy params struct */
|
4275 |
|
|
|
4276 |
|
|
/* Initialize pointers */
|
4277 |
|
|
bond->first_slave = NULL;
|
4278 |
|
|
bond->curr_active_slave = NULL;
|
4279 |
|
|
bond->current_arp_slave = NULL;
|
4280 |
|
|
bond->primary_slave = NULL;
|
4281 |
|
|
bond->dev = bond_dev;
|
4282 |
|
|
INIT_LIST_HEAD(&bond->vlan_list);
|
4283 |
|
|
|
4284 |
|
|
/* Initialize the device entry points */
|
4285 |
|
|
bond_dev->open = bond_open;
|
4286 |
|
|
bond_dev->stop = bond_close;
|
4287 |
|
|
bond_dev->get_stats = bond_get_stats;
|
4288 |
|
|
bond_dev->do_ioctl = bond_do_ioctl;
|
4289 |
|
|
bond_dev->set_multicast_list = bond_set_multicast_list;
|
4290 |
|
|
bond_dev->change_mtu = bond_change_mtu;
|
4291 |
|
|
bond_dev->set_mac_address = bond_set_mac_address;
|
4292 |
|
|
|
4293 |
|
|
bond_set_mode_ops(bond_dev, bond->params.mode);
|
4294 |
|
|
|
4295 |
|
|
#ifdef CONFIG_NET_FASTROUTE
|
4296 |
|
|
bond_dev->accept_fastpath = bond_accept_fastpath;
|
4297 |
|
|
#endif
|
4298 |
|
|
|
4299 |
|
|
/* Initialize the device options */
|
4300 |
|
|
bond_dev->tx_queue_len = 0;
|
4301 |
|
|
bond_dev->flags |= IFF_MASTER|IFF_MULTICAST;
|
4302 |
|
|
|
4303 |
|
|
/* At first, we block adding VLANs. That's the only way to
|
4304 |
|
|
* prevent problems that occur when adding VLANs over an
|
4305 |
|
|
* empty bond. The block will be removed once non-challenged
|
4306 |
|
|
* slaves are enslaved.
|
4307 |
|
|
*/
|
4308 |
|
|
bond_dev->features |= NETIF_F_VLAN_CHALLENGED;
|
4309 |
|
|
|
4310 |
|
|
/* By default, we declare the bond to be fully
|
4311 |
|
|
* VLAN hardware accelerated capable. Special
|
4312 |
|
|
* care is taken in the various xmit functions
|
4313 |
|
|
* when there are slaves that are not hw accel
|
4314 |
|
|
* capable
|
4315 |
|
|
*/
|
4316 |
|
|
bond_dev->vlan_rx_register = bond_vlan_rx_register;
|
4317 |
|
|
bond_dev->vlan_rx_add_vid = bond_vlan_rx_add_vid;
|
4318 |
|
|
bond_dev->vlan_rx_kill_vid = bond_vlan_rx_kill_vid;
|
4319 |
|
|
bond_dev->features |= (NETIF_F_HW_VLAN_TX |
|
4320 |
|
|
NETIF_F_HW_VLAN_RX |
|
4321 |
|
|
NETIF_F_HW_VLAN_FILTER);
|
4322 |
|
|
|
4323 |
|
|
#ifdef CONFIG_PROC_FS
|
4324 |
|
|
bond_create_proc_entry(bond);
|
4325 |
|
|
#endif
|
4326 |
|
|
|
4327 |
|
|
list_add_tail(&bond->bond_list, &bond_dev_list);
|
4328 |
|
|
|
4329 |
|
|
return 0;
|
4330 |
|
|
}
|
4331 |
|
|
|
4332 |
|
|
/* De-initialize device specific data.
|
4333 |
|
|
* Caller must hold rtnl_lock.
|
4334 |
|
|
*/
|
4335 |
|
|
static inline void bond_deinit(struct net_device *bond_dev)
|
4336 |
|
|
{
|
4337 |
|
|
struct bonding *bond = bond_dev->priv;
|
4338 |
|
|
|
4339 |
|
|
list_del(&bond->bond_list);
|
4340 |
|
|
|
4341 |
|
|
#ifdef CONFIG_PROC_FS
|
4342 |
|
|
bond_remove_proc_entry(bond);
|
4343 |
|
|
#endif
|
4344 |
|
|
}
|
4345 |
|
|
|
4346 |
|
|
/* Unregister and free all bond devices.
|
4347 |
|
|
* Caller must hold rtnl_lock.
|
4348 |
|
|
*/
|
4349 |
|
|
static void bond_free_all(void)
|
4350 |
|
|
{
|
4351 |
|
|
struct bonding *bond, *nxt;
|
4352 |
|
|
|
4353 |
|
|
list_for_each_entry_safe(bond, nxt, &bond_dev_list, bond_list) {
|
4354 |
|
|
struct net_device *bond_dev = bond->dev;
|
4355 |
|
|
|
4356 |
|
|
unregister_netdevice(bond_dev);
|
4357 |
|
|
bond_deinit(bond_dev);
|
4358 |
|
|
free_netdev(bond_dev);
|
4359 |
|
|
}
|
4360 |
|
|
|
4361 |
|
|
#ifdef CONFIG_PROC_FS
|
4362 |
|
|
bond_destroy_proc_dir();
|
4363 |
|
|
#endif
|
4364 |
|
|
}
|
4365 |
|
|
|
4366 |
|
|
/*------------------------- Module initialization ---------------------------*/
|
4367 |
|
|
|
4368 |
|
|
/*
|
4369 |
|
|
* Convert string input module parms. Accept either the
|
4370 |
|
|
* number of the mode or its string name.
|
4371 |
|
|
*/
|
4372 |
|
|
static inline int bond_parse_parm(char *mode_arg, struct bond_parm_tbl *tbl)
|
4373 |
|
|
{
|
4374 |
|
|
int i;
|
4375 |
|
|
|
4376 |
|
|
for (i = 0; tbl[i].modename; i++) {
|
4377 |
|
|
if ((isdigit(*mode_arg) &&
|
4378 |
|
|
tbl[i].mode == simple_strtol(mode_arg, NULL, 0)) ||
|
4379 |
|
|
(strncmp(mode_arg, tbl[i].modename,
|
4380 |
|
|
strlen(tbl[i].modename)) == 0)) {
|
4381 |
|
|
return tbl[i].mode;
|
4382 |
|
|
}
|
4383 |
|
|
}
|
4384 |
|
|
|
4385 |
|
|
return -1;
|
4386 |
|
|
}
|
4387 |
|
|
|
4388 |
|
|
static int bond_check_params(struct bond_params *params)
|
4389 |
|
|
{
|
4390 |
|
|
/*
|
4391 |
|
|
* Convert string parameters.
|
4392 |
|
|
*/
|
4393 |
|
|
if (mode) {
|
4394 |
|
|
bond_mode = bond_parse_parm(mode, bond_mode_tbl);
|
4395 |
|
|
if (bond_mode == -1) {
|
4396 |
|
|
printk(KERN_ERR DRV_NAME
|
4397 |
|
|
": Error: Invalid bonding mode \"%s\"\n",
|
4398 |
|
|
mode == NULL ? "NULL" : mode);
|
4399 |
|
|
return -EINVAL;
|
4400 |
|
|
}
|
4401 |
|
|
}
|
4402 |
|
|
|
4403 |
|
|
if (lacp_rate) {
|
4404 |
|
|
if (bond_mode != BOND_MODE_8023AD) {
|
4405 |
|
|
printk(KERN_INFO DRV_NAME
|
4406 |
|
|
": lacp_rate param is irrelevant in mode %s\n",
|
4407 |
|
|
bond_mode_name(bond_mode));
|
4408 |
|
|
} else {
|
4409 |
|
|
lacp_fast = bond_parse_parm(lacp_rate, bond_lacp_tbl);
|
4410 |
|
|
if (lacp_fast == -1) {
|
4411 |
|
|
printk(KERN_ERR DRV_NAME
|
4412 |
|
|
": Error: Invalid lacp rate \"%s\"\n",
|
4413 |
|
|
lacp_rate == NULL ? "NULL" : lacp_rate);
|
4414 |
|
|
return -EINVAL;
|
4415 |
|
|
}
|
4416 |
|
|
}
|
4417 |
|
|
}
|
4418 |
|
|
|
4419 |
|
|
if (max_bonds < 1 || max_bonds > INT_MAX) {
|
4420 |
|
|
printk(KERN_WARNING DRV_NAME
|
4421 |
|
|
": Warning: max_bonds (%d) not in range %d-%d, so it "
|
4422 |
|
|
"was reset to BOND_DEFAULT_MAX_BONDS (%d)",
|
4423 |
|
|
max_bonds, 1, INT_MAX, BOND_DEFAULT_MAX_BONDS);
|
4424 |
|
|
max_bonds = BOND_DEFAULT_MAX_BONDS;
|
4425 |
|
|
}
|
4426 |
|
|
|
4427 |
|
|
if (miimon < 0) {
|
4428 |
|
|
printk(KERN_WARNING DRV_NAME
|
4429 |
|
|
": Warning: miimon module parameter (%d), "
|
4430 |
|
|
"not in range 0-%d, so it was reset to %d\n",
|
4431 |
|
|
miimon, INT_MAX, BOND_LINK_MON_INTERV);
|
4432 |
|
|
miimon = BOND_LINK_MON_INTERV;
|
4433 |
|
|
}
|
4434 |
|
|
|
4435 |
|
|
if (updelay < 0) {
|
4436 |
|
|
printk(KERN_WARNING DRV_NAME
|
4437 |
|
|
": Warning: updelay module parameter (%d), "
|
4438 |
|
|
"not in range 0-%d, so it was reset to 0\n",
|
4439 |
|
|
updelay, INT_MAX);
|
4440 |
|
|
updelay = 0;
|
4441 |
|
|
}
|
4442 |
|
|
|
4443 |
|
|
if (downdelay < 0) {
|
4444 |
|
|
printk(KERN_WARNING DRV_NAME
|
4445 |
|
|
": Warning: downdelay module parameter (%d), "
|
4446 |
|
|
"not in range 0-%d, so it was reset to 0\n",
|
4447 |
|
|
downdelay, INT_MAX);
|
4448 |
|
|
downdelay = 0;
|
4449 |
|
|
}
|
4450 |
|
|
|
4451 |
|
|
if ((use_carrier != 0) && (use_carrier != 1)) {
|
4452 |
|
|
printk(KERN_WARNING DRV_NAME
|
4453 |
|
|
": Warning: use_carrier module parameter (%d), "
|
4454 |
|
|
"not of valid value (0/1), so it was set to 1\n",
|
4455 |
|
|
use_carrier);
|
4456 |
|
|
use_carrier = 1;
|
4457 |
|
|
}
|
4458 |
|
|
|
4459 |
|
|
/* reset values for 802.3ad */
|
4460 |
|
|
if (bond_mode == BOND_MODE_8023AD) {
|
4461 |
|
|
if (!miimon) {
|
4462 |
|
|
printk(KERN_WARNING DRV_NAME
|
4463 |
|
|
": Warning: miimon must be specified, "
|
4464 |
|
|
"otherwise bonding will not detect link "
|
4465 |
|
|
"failure, speed and duplex which are "
|
4466 |
|
|
"essential for 802.3ad operation\n");
|
4467 |
|
|
printk(KERN_WARNING "Forcing miimon to 100msec\n");
|
4468 |
|
|
miimon = 100;
|
4469 |
|
|
}
|
4470 |
|
|
}
|
4471 |
|
|
|
4472 |
|
|
/* reset values for TLB/ALB */
|
4473 |
|
|
if ((bond_mode == BOND_MODE_TLB) ||
|
4474 |
|
|
(bond_mode == BOND_MODE_ALB)) {
|
4475 |
|
|
if (!miimon) {
|
4476 |
|
|
printk(KERN_WARNING DRV_NAME
|
4477 |
|
|
": Warning: miimon must be specified, "
|
4478 |
|
|
"otherwise bonding will not detect link "
|
4479 |
|
|
"failure and link speed which are essential "
|
4480 |
|
|
"for TLB/ALB load balancing\n");
|
4481 |
|
|
printk(KERN_WARNING "Forcing miimon to 100msec\n");
|
4482 |
|
|
miimon = 100;
|
4483 |
|
|
}
|
4484 |
|
|
}
|
4485 |
|
|
|
4486 |
|
|
if (bond_mode == BOND_MODE_ALB) {
|
4487 |
|
|
printk(KERN_NOTICE DRV_NAME
|
4488 |
|
|
": In ALB mode you might experience client "
|
4489 |
|
|
"disconnections upon reconnection of a link if the "
|
4490 |
|
|
"bonding module updelay parameter (%d msec) is "
|
4491 |
|
|
"incompatible with the forwarding delay time of the "
|
4492 |
|
|
"switch\n",
|
4493 |
|
|
updelay);
|
4494 |
|
|
}
|
4495 |
|
|
|
4496 |
|
|
if (!miimon) {
|
4497 |
|
|
if (updelay || downdelay) {
|
4498 |
|
|
/* just warn the user the up/down delay will have
|
4499 |
|
|
* no effect since miimon is zero...
|
4500 |
|
|
*/
|
4501 |
|
|
printk(KERN_WARNING DRV_NAME
|
4502 |
|
|
": Warning: miimon module parameter not set "
|
4503 |
|
|
"and updelay (%d) or downdelay (%d) module "
|
4504 |
|
|
"parameter is set; updelay and downdelay have "
|
4505 |
|
|
"no effect unless miimon is set\n",
|
4506 |
|
|
updelay, downdelay);
|
4507 |
|
|
}
|
4508 |
|
|
} else {
|
4509 |
|
|
/* don't allow arp monitoring */
|
4510 |
|
|
if (arp_interval) {
|
4511 |
|
|
printk(KERN_WARNING DRV_NAME
|
4512 |
|
|
": Warning: miimon (%d) and arp_interval (%d) "
|
4513 |
|
|
"can't be used simultaneously, disabling ARP "
|
4514 |
|
|
"monitoring\n",
|
4515 |
|
|
miimon, arp_interval);
|
4516 |
|
|
arp_interval = 0;
|
4517 |
|
|
}
|
4518 |
|
|
|
4519 |
|
|
if ((updelay % miimon) != 0) {
|
4520 |
|
|
printk(KERN_WARNING DRV_NAME
|
4521 |
|
|
": Warning: updelay (%d) is not a multiple "
|
4522 |
|
|
"of miimon (%d), updelay rounded to %d ms\n",
|
4523 |
|
|
updelay, miimon, (updelay / miimon) * miimon);
|
4524 |
|
|
}
|
4525 |
|
|
|
4526 |
|
|
updelay /= miimon;
|
4527 |
|
|
|
4528 |
|
|
if ((downdelay % miimon) != 0) {
|
4529 |
|
|
printk(KERN_WARNING DRV_NAME
|
4530 |
|
|
": Warning: downdelay (%d) is not a multiple "
|
4531 |
|
|
"of miimon (%d), downdelay rounded to %d ms\n",
|
4532 |
|
|
downdelay, miimon,
|
4533 |
|
|
(downdelay / miimon) * miimon);
|
4534 |
|
|
}
|
4535 |
|
|
|
4536 |
|
|
downdelay /= miimon;
|
4537 |
|
|
}
|
4538 |
|
|
|
4539 |
|
|
if (arp_interval < 0) {
|
4540 |
|
|
printk(KERN_WARNING DRV_NAME
|
4541 |
|
|
": Warning: arp_interval module parameter (%d) "
|
4542 |
|
|
", not in range 0-%d, so it was reset to %d\n",
|
4543 |
|
|
arp_interval, INT_MAX, BOND_LINK_ARP_INTERV);
|
4544 |
|
|
arp_interval = BOND_LINK_ARP_INTERV;
|
4545 |
|
|
}
|
4546 |
|
|
|
4547 |
|
|
for (arp_ip_count = 0;
|
4548 |
|
|
(arp_ip_count < BOND_MAX_ARP_TARGETS) && arp_ip_target[arp_ip_count];
|
4549 |
|
|
arp_ip_count++) {
|
4550 |
|
|
/* not complete check, but should be good enough to
|
4551 |
|
|
catch mistakes */
|
4552 |
|
|
if (!isdigit(arp_ip_target[arp_ip_count][0])) {
|
4553 |
|
|
printk(KERN_WARNING DRV_NAME
|
4554 |
|
|
": Warning: bad arp_ip_target module parameter "
|
4555 |
|
|
"(%s), ARP monitoring will not be performed\n",
|
4556 |
|
|
arp_ip_target[arp_ip_count]);
|
4557 |
|
|
arp_interval = 0;
|
4558 |
|
|
} else {
|
4559 |
|
|
u32 ip = in_aton(arp_ip_target[arp_ip_count]);
|
4560 |
|
|
arp_target[arp_ip_count] = ip;
|
4561 |
|
|
}
|
4562 |
|
|
}
|
4563 |
|
|
|
4564 |
|
|
if (arp_interval && !arp_ip_count) {
|
4565 |
|
|
/* don't allow arping if no arp_ip_target given... */
|
4566 |
|
|
printk(KERN_WARNING DRV_NAME
|
4567 |
|
|
": Warning: arp_interval module parameter (%d) "
|
4568 |
|
|
"specified without providing an arp_ip_target "
|
4569 |
|
|
"parameter, arp_interval was reset to 0\n",
|
4570 |
|
|
arp_interval);
|
4571 |
|
|
arp_interval = 0;
|
4572 |
|
|
}
|
4573 |
|
|
|
4574 |
|
|
if (miimon) {
|
4575 |
|
|
printk(KERN_INFO DRV_NAME
|
4576 |
|
|
": MII link monitoring set to %d ms\n",
|
4577 |
|
|
miimon);
|
4578 |
|
|
} else if (arp_interval) {
|
4579 |
|
|
int i;
|
4580 |
|
|
|
4581 |
|
|
printk(KERN_INFO DRV_NAME
|
4582 |
|
|
": ARP monitoring set to %d ms with %d target(s):",
|
4583 |
|
|
arp_interval, arp_ip_count);
|
4584 |
|
|
|
4585 |
|
|
for (i = 0; i < arp_ip_count; i++)
|
4586 |
|
|
printk (" %s", arp_ip_target[i]);
|
4587 |
|
|
|
4588 |
|
|
printk("\n");
|
4589 |
|
|
|
4590 |
|
|
} else {
|
4591 |
|
|
/* miimon and arp_interval not set, we need one so things
|
4592 |
|
|
* work as expected, see bonding.txt for details
|
4593 |
|
|
*/
|
4594 |
|
|
printk(KERN_WARNING DRV_NAME
|
4595 |
|
|
": Warning: either miimon or arp_interval and "
|
4596 |
|
|
"arp_ip_target module parameters must be specified, "
|
4597 |
|
|
"otherwise bonding will not detect link failures! see "
|
4598 |
|
|
"bonding.txt for details.\n");
|
4599 |
|
|
}
|
4600 |
|
|
|
4601 |
|
|
if (primary && !USES_PRIMARY(bond_mode)) {
|
4602 |
|
|
/* currently, using a primary only makes sense
|
4603 |
|
|
* in active backup, TLB or ALB modes
|
4604 |
|
|
*/
|
4605 |
|
|
printk(KERN_WARNING DRV_NAME
|
4606 |
|
|
": Warning: %s primary device specified but has no "
|
4607 |
|
|
"effect in %s mode\n",
|
4608 |
|
|
primary, bond_mode_name(bond_mode));
|
4609 |
|
|
primary = NULL;
|
4610 |
|
|
}
|
4611 |
|
|
|
4612 |
|
|
/* fill params struct with the proper values */
|
4613 |
|
|
params->mode = bond_mode;
|
4614 |
|
|
params->miimon = miimon;
|
4615 |
|
|
params->arp_interval = arp_interval;
|
4616 |
|
|
params->updelay = updelay;
|
4617 |
|
|
params->downdelay = downdelay;
|
4618 |
|
|
params->use_carrier = use_carrier;
|
4619 |
|
|
params->lacp_fast = lacp_fast;
|
4620 |
|
|
params->primary[0] = 0;
|
4621 |
|
|
|
4622 |
|
|
if (primary) {
|
4623 |
|
|
strncpy(params->primary, primary, IFNAMSIZ);
|
4624 |
|
|
params->primary[IFNAMSIZ - 1] = 0;
|
4625 |
|
|
}
|
4626 |
|
|
|
4627 |
|
|
memcpy(params->arp_targets, arp_target, sizeof(arp_target));
|
4628 |
|
|
|
4629 |
|
|
return 0;
|
4630 |
|
|
}
|
4631 |
|
|
|
4632 |
|
|
static int __init bonding_init(void)
|
4633 |
|
|
{
|
4634 |
|
|
struct bond_params params;
|
4635 |
|
|
int i;
|
4636 |
|
|
int res;
|
4637 |
|
|
|
4638 |
|
|
printk(KERN_INFO "%s", version);
|
4639 |
|
|
|
4640 |
|
|
res = bond_check_params(¶ms);
|
4641 |
|
|
if (res) {
|
4642 |
|
|
return res;
|
4643 |
|
|
}
|
4644 |
|
|
|
4645 |
|
|
rtnl_lock();
|
4646 |
|
|
|
4647 |
|
|
#ifdef CONFIG_PROC_FS
|
4648 |
|
|
bond_create_proc_dir();
|
4649 |
|
|
#endif
|
4650 |
|
|
|
4651 |
|
|
for (i = 0; i < max_bonds; i++) {
|
4652 |
|
|
struct net_device *bond_dev;
|
4653 |
|
|
|
4654 |
|
|
bond_dev = alloc_netdev(sizeof(struct bonding), "", ether_setup);
|
4655 |
|
|
if (!bond_dev) {
|
4656 |
|
|
res = -ENOMEM;
|
4657 |
|
|
goto out_err;
|
4658 |
|
|
}
|
4659 |
|
|
|
4660 |
|
|
res = dev_alloc_name(bond_dev, "bond%d");
|
4661 |
|
|
if (res < 0) {
|
4662 |
|
|
free_netdev(bond_dev);
|
4663 |
|
|
goto out_err;
|
4664 |
|
|
}
|
4665 |
|
|
|
4666 |
|
|
/* bond_init() must be called after dev_alloc_name() (for the
|
4667 |
|
|
* /proc files), but before register_netdevice(), because we
|
4668 |
|
|
* need to set function pointers.
|
4669 |
|
|
*/
|
4670 |
|
|
res = bond_init(bond_dev, ¶ms);
|
4671 |
|
|
if (res < 0) {
|
4672 |
|
|
free_netdev(bond_dev);
|
4673 |
|
|
goto out_err;
|
4674 |
|
|
}
|
4675 |
|
|
|
4676 |
|
|
SET_MODULE_OWNER(bond_dev);
|
4677 |
|
|
|
4678 |
|
|
res = register_netdevice(bond_dev);
|
4679 |
|
|
if (res < 0) {
|
4680 |
|
|
bond_deinit(bond_dev);
|
4681 |
|
|
free_netdev(bond_dev);
|
4682 |
|
|
goto out_err;
|
4683 |
|
|
}
|
4684 |
|
|
}
|
4685 |
|
|
|
4686 |
|
|
rtnl_unlock();
|
4687 |
|
|
register_netdevice_notifier(&bond_netdev_notifier);
|
4688 |
|
|
|
4689 |
|
|
return 0;
|
4690 |
|
|
|
4691 |
|
|
out_err:
|
4692 |
|
|
/* free and unregister all bonds that were successfully added */
|
4693 |
|
|
bond_free_all();
|
4694 |
|
|
|
4695 |
|
|
rtnl_unlock();
|
4696 |
|
|
|
4697 |
|
|
return res;
|
4698 |
|
|
}
|
4699 |
|
|
|
4700 |
|
|
static void __exit bonding_exit(void)
|
4701 |
|
|
{
|
4702 |
|
|
unregister_netdevice_notifier(&bond_netdev_notifier);
|
4703 |
|
|
|
4704 |
|
|
rtnl_lock();
|
4705 |
|
|
bond_free_all();
|
4706 |
|
|
rtnl_unlock();
|
4707 |
|
|
}
|
4708 |
|
|
|
4709 |
|
|
module_init(bonding_init);
|
4710 |
|
|
module_exit(bonding_exit);
|
4711 |
|
|
MODULE_LICENSE("GPL");
|
4712 |
|
|
MODULE_DESCRIPTION(DRV_DESCRIPTION ", v" DRV_VERSION);
|
4713 |
|
|
MODULE_AUTHOR("Thomas Davis, tadavis@lbl.gov and many others");
|
4714 |
|
|
MODULE_SUPPORTED_DEVICE("most ethernet devices");
|
4715 |
|
|
|
4716 |
|
|
/*
|
4717 |
|
|
* Local variables:
|
4718 |
|
|
* c-indent-level: 8
|
4719 |
|
|
* c-basic-offset: 8
|
4720 |
|
|
* tab-width: 8
|
4721 |
|
|
* End:
|
4722 |
|
|
*/
|
4723 |
|
|
|