blob: 5ef1047e2e6635b4cf060fa2f927820d917f5dc4 [file] [log] [blame]
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -03001=====================
2The Linux IPMI Driver
3=====================
Linus Torvalds1da177e2005-04-16 15:20:36 -07004
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -03005:Author: Corey Minyard <minyard@mvista.com> / <minyard@acm.org>
Linus Torvalds1da177e2005-04-16 15:20:36 -07006
7The Intelligent Platform Management Interface, or IPMI, is a
8standard for controlling intelligent devices that monitor a system.
9It provides for dynamic discovery of sensors in the system and the
10ability to monitor the sensors and be informed when the sensor's
11values change or go outside certain boundaries. It also has a
Matt LaPlantedc474c82006-06-30 01:56:06 -070012standardized database for field-replaceable units (FRUs) and a watchdog
Linus Torvalds1da177e2005-04-16 15:20:36 -070013timer.
14
15To use this, you need an interface to an IPMI controller in your
16system (called a Baseboard Management Controller, or BMC) and
17management software that can use the IPMI system.
18
19This document describes how to use the IPMI driver for Linux. If you
20are not familiar with IPMI itself, see the web site at
21http://www.intel.com/design/servers/ipmi/index.htm. IPMI is a big
22subject and I can't cover it all here!
23
24Configuration
25-------------
26
Corey Minyard845e78a2005-06-23 22:01:39 -070027The Linux IPMI driver is modular, which means you have to pick several
Linus Torvalds1da177e2005-04-16 15:20:36 -070028things to have it work right depending on your hardware. Most of
Corey Minyard845e78a2005-06-23 22:01:39 -070029these are available in the 'Character Devices' menu then the IPMI
30menu.
Linus Torvalds1da177e2005-04-16 15:20:36 -070031
32No matter what, you must pick 'IPMI top-level message handler' to use
33IPMI. What you do beyond that depends on your needs and hardware.
34
35The message handler does not provide any user-level interfaces.
36Kernel code (like the watchdog) can still use it. If you need access
37from userland, you need to select 'Device interface for IPMI' if you
Corey Minyard845e78a2005-06-23 22:01:39 -070038want access through a device driver.
Linus Torvalds1da177e2005-04-16 15:20:36 -070039
Corey Minyard845e78a2005-06-23 22:01:39 -070040The driver interface depends on your hardware. If your system
41properly provides the SMBIOS info for IPMI, the driver will detect it
42and just work. If you have a board with a standard interface (These
43will generally be either "KCS", "SMIC", or "BT", consult your hardware
Corey Minyard25930702012-03-19 16:00:55 -050044manual), choose the 'IPMI SI handler' option. A driver also exists
45for direct I2C access to the IPMI management controller. Some boards
46support this, but it is unknown if it will work on every board. For
47this, choose 'IPMI SMBus handler', but be ready to try to do some
48figuring to see if it will work on your system if the SMBIOS/APCI
49information is wrong or not present. It is fairly safe to have both
50these enabled and let the drivers auto-detect what is present.
Linus Torvalds1da177e2005-04-16 15:20:36 -070051
52You should generally enable ACPI on your system, as systems with IPMI
Corey Minyard845e78a2005-06-23 22:01:39 -070053can have ACPI tables describing them.
Linus Torvalds1da177e2005-04-16 15:20:36 -070054
55If you have a standard interface and the board manufacturer has done
56their job correctly, the IPMI controller should be automatically
Corey Minyard845e78a2005-06-23 22:01:39 -070057detected (via ACPI or SMBIOS tables) and should just work. Sadly,
58many boards do not have this information. The driver attempts
59standard defaults, but they may not work. If you fall into this
Corey Minyard25930702012-03-19 16:00:55 -050060situation, you need to read the section below named 'The SI Driver' or
61"The SMBus Driver" on how to hand-configure your system.
Linus Torvalds1da177e2005-04-16 15:20:36 -070062
63IPMI defines a standard watchdog timer. You can enable this with the
64'IPMI Watchdog Timer' config option. If you compile the driver into
65the kernel, then via a kernel command-line option you can have the
Matt LaPlantedc474c82006-06-30 01:56:06 -070066watchdog timer start as soon as it initializes. It also have a lot
Linus Torvalds1da177e2005-04-16 15:20:36 -070067of other options, see the 'Watchdog' section below for more details.
68Note that you can also have the watchdog continue to run if it is
69closed (by default it is disabled on close). Go into the 'Watchdog
70Cards' menu, enable 'Watchdog Timer Support', and enable the option
71'Disable watchdog shutdown on close'.
72
Corey Minyard845e78a2005-06-23 22:01:39 -070073IPMI systems can often be powered off using IPMI commands. Select
74'IPMI Poweroff' to do this. The driver will auto-detect if the system
75can be powered off by IPMI. It is safe to enable this even if your
76system doesn't support this option. This works on ATCA systems, the
77Radisys CPI1 card, and any IPMI system that supports standard chassis
78management commands.
79
80If you want the driver to put an event into the event log on a panic,
81enable the 'Generate a panic event to all BMCs on a panic' option. If
82you want the whole panic string put into the event log using OEM
83events, enable the 'Generate OEM events containing the panic string'
Corey Minyard1c9f98d2017-08-18 17:32:03 -050084option. You can also enable these dynamically by setting the module
85parameter named "panic_op" in the ipmi_msghandler module to "event"
86or "string". Setting that parameter to "none" disables this function.
Linus Torvalds1da177e2005-04-16 15:20:36 -070087
88Basic Design
89------------
90
91The Linux IPMI driver is designed to be very modular and flexible, you
92only need to take the pieces you need and you can use it in many
93different ways. Because of that, it's broken into many chunks of
Corey Minyard845e78a2005-06-23 22:01:39 -070094code. These chunks (by module name) are:
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
96ipmi_msghandler - This is the central piece of software for the IPMI
97system. It handles all messages, message timing, and responses. The
98IPMI users tie into this, and the IPMI physical interfaces (called
99System Management Interfaces, or SMIs) also tie in here. This
100provides the kernelland interface for IPMI, but does not provide an
101interface for use by application processes.
102
103ipmi_devintf - This provides a userland IOCTL interface for the IPMI
104driver, each open file for this device ties in to the message handler
105as an IPMI user.
106
Corey Minyard845e78a2005-06-23 22:01:39 -0700107ipmi_si - A driver for various system interfaces. This supports KCS,
Corey Minyard25930702012-03-19 16:00:55 -0500108SMIC, and BT interfaces. Unless you have an SMBus interface or your
109own custom interface, you probably need to use this.
110
111ipmi_ssif - A driver for accessing BMCs on the SMBus. It uses the
112I2C kernel driver's SMBus interfaces to send and receive IPMI messages
113over the SMBus.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114
Corey Minyardc11daf62016-10-25 20:31:53 -0500115ipmi_powernv - A driver for access BMCs on POWERNV systems.
116
Corey Minyard845e78a2005-06-23 22:01:39 -0700117ipmi_watchdog - IPMI requires systems to have a very capable watchdog
118timer. This driver implements the standard Linux watchdog timer
119interface on top of the IPMI message handler.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120
Corey Minyard845e78a2005-06-23 22:01:39 -0700121ipmi_poweroff - Some systems support the ability to be turned off via
122IPMI commands.
123
Corey Minyardc11daf62016-10-25 20:31:53 -0500124bt-bmc - This is not part of the main driver, but instead a driver for
125accessing a BMC-side interface of a BT interface. It is used on BMCs
126running Linux to provide an interface to the host.
Corey Minyard845e78a2005-06-23 22:01:39 -0700127
Corey Minyardc11daf62016-10-25 20:31:53 -0500128These are all individually selectable via configuration options.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129
130Much documentation for the interface is in the include files. The
131IPMI include files are:
132
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133linux/ipmi.h - Contains the user interface and IOCTL interface for IPMI.
134
135linux/ipmi_smi.h - Contains the interface for system management interfaces
136(things that interface to IPMI controllers) to use.
137
138linux/ipmi_msgdefs.h - General definitions for base IPMI messaging.
139
140
141Addressing
142----------
143
144The IPMI addressing works much like IP addresses, you have an overlay
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300145to handle the different address types. The overlay is::
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146
147 struct ipmi_addr
148 {
149 int addr_type;
150 short channel;
151 char data[IPMI_MAX_ADDR_SIZE];
152 };
153
154The addr_type determines what the address really is. The driver
155currently understands two different types of addresses.
156
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300157"System Interface" addresses are defined as::
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158
159 struct ipmi_system_interface_addr
160 {
161 int addr_type;
162 short channel;
163 };
164
165and the type is IPMI_SYSTEM_INTERFACE_ADDR_TYPE. This is used for talking
166straight to the BMC on the current card. The channel must be
167IPMI_BMC_CHANNEL.
168
169Messages that are destined to go out on the IPMB bus use the
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300170IPMI_IPMB_ADDR_TYPE address type. The format is::
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171
172 struct ipmi_ipmb_addr
173 {
174 int addr_type;
175 short channel;
176 unsigned char slave_addr;
177 unsigned char lun;
178 };
179
180The "channel" here is generally zero, but some devices support more
181than one channel, it corresponds to the channel as defined in the IPMI
182spec.
183
184
185Messages
186--------
187
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300188Messages are defined as::
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300190 struct ipmi_msg
191 {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192 unsigned char netfn;
193 unsigned char lun;
194 unsigned char cmd;
195 unsigned char *data;
196 int data_len;
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300197 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700198
199The driver takes care of adding/stripping the header information. The
200data portion is just the data to be send (do NOT put addressing info
201here) or the response. Note that the completion code of a response is
202the first item in "data", it is not stripped out because that is how
203all the messages are defined in the spec (and thus makes counting the
204offsets a little easier :-).
205
206When using the IOCTL interface from userland, you must provide a block
207of data for "data", fill it, and set data_len to the length of the
208block of data, even when receiving messages. Otherwise the driver
209will have no place to put the message.
210
211Messages coming up from the message handler in kernelland will come in
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300212as::
Linus Torvalds1da177e2005-04-16 15:20:36 -0700213
214 struct ipmi_recv_msg
215 {
216 struct list_head link;
217
218 /* The type of message as defined in the "Receive Types"
219 defines above. */
220 int recv_type;
221
222 ipmi_user_t *user;
223 struct ipmi_addr addr;
224 long msgid;
225 struct ipmi_msg msg;
226
227 /* Call this when done with the message. It will presumably free
228 the message and do any other necessary cleanup. */
229 void (*done)(struct ipmi_recv_msg *msg);
230
231 /* Place-holder for the data, don't make any assumptions about
232 the size or existence of this, since it may change. */
233 unsigned char msg_data[IPMI_MAX_MSG_LENGTH];
234 };
235
236You should look at the receive type and handle the message
237appropriately.
238
239
240The Upper Layer Interface (Message Handler)
241-------------------------------------------
242
243The upper layer of the interface provides the users with a consistent
244view of the IPMI interfaces. It allows multiple SMI interfaces to be
245addressed (because some boards actually have multiple BMCs on them)
246and the user should not have to care what type of SMI is below them.
247
248
Corey Minyardc11daf62016-10-25 20:31:53 -0500249Watching For Interfaces
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300250^^^^^^^^^^^^^^^^^^^^^^^
Corey Minyardc11daf62016-10-25 20:31:53 -0500251
252When your code comes up, the IPMI driver may or may not have detected
253if IPMI devices exist. So you might have to defer your setup until
254the device is detected, or you might be able to do it immediately.
255To handle this, and to allow for discovery, you register an SMI
256watcher with ipmi_smi_watcher_register() to iterate over interfaces
257and tell you when they come and go.
258
259
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260Creating the User
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300261^^^^^^^^^^^^^^^^^
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262
Rami Rosen08f6cd02016-12-18 21:11:54 +0200263To use the message handler, you must first create a user using
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264ipmi_create_user. The interface number specifies which SMI you want
265to connect to, and you must supply callback functions to be called
266when data comes in. The callback function can run at interrupt level,
267so be careful using the callbacks. This also allows to you pass in a
268piece of data, the handler_data, that will be passed back to you on
269all calls.
270
271Once you are done, call ipmi_destroy_user() to get rid of the user.
272
273From userland, opening the device automatically creates a user, and
274closing the device automatically destroys the user.
275
276
277Messaging
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300278^^^^^^^^^
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279
Corey Minyardc11daf62016-10-25 20:31:53 -0500280To send a message from kernel-land, the ipmi_request_settime() call does
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281pretty much all message handling. Most of the parameter are
282self-explanatory. However, it takes a "msgid" parameter. This is NOT
283the sequence number of messages. It is simply a long value that is
284passed back when the response for the message is returned. You may
285use it for anything you like.
286
287Responses come back in the function pointed to by the ipmi_recv_hndl
288field of the "handler" that you passed in to ipmi_create_user().
289Remember again, these may be running at interrupt level. Remember to
290look at the receive type, too.
291
292From userland, you fill out an ipmi_req_t structure and use the
293IPMICTL_SEND_COMMAND ioctl. For incoming stuff, you can use select()
294or poll() to wait for messages to come in. However, you cannot use
295read() to get them, you must call the IPMICTL_RECEIVE_MSG with the
296ipmi_recv_t structure to actually get the message. Remember that you
297must supply a pointer to a block of data in the msg.data field, and
298you must fill in the msg.data_len field with the size of the data.
299This gives the receiver a place to actually put the message.
300
301If the message cannot fit into the data you provide, you will get an
302EMSGSIZE error and the driver will leave the data in the receive
303queue. If you want to get it and have it truncate the message, us
304the IPMICTL_RECEIVE_MSG_TRUNC ioctl.
305
306When you send a command (which is defined by the lowest-order bit of
307the netfn per the IPMI spec) on the IPMB bus, the driver will
308automatically assign the sequence number to the command and save the
309command. If the response is not receive in the IPMI-specified 5
310seconds, it will generate a response automatically saying the command
311timed out. If an unsolicited response comes in (if it was after 5
312seconds, for instance), that response will be ignored.
313
314In kernelland, after you receive a message and are done with it, you
315MUST call ipmi_free_recv_msg() on it, or you will leak messages. Note
316that you should NEVER mess with the "done" field of a message, that is
317required to properly clean up the message.
318
319Note that when sending, there is an ipmi_request_supply_msgs() call
320that lets you supply the smi and receive message. This is useful for
321pieces of code that need to work even if the system is out of buffers
322(the watchdog timer uses this, for instance). You supply your own
323buffer and own free routines. This is not recommended for normal use,
324though, since it is tricky to manage your own buffers.
325
326
327Events and Incoming Commands
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300328^^^^^^^^^^^^^^^^^^^^^^^^^^^^
Linus Torvalds1da177e2005-04-16 15:20:36 -0700329
330The driver takes care of polling for IPMI events and receiving
331commands (commands are messages that are not responses, they are
332commands that other things on the IPMB bus have sent you). To receive
333these, you must register for them, they will not automatically be sent
334to you.
335
336To receive events, you must call ipmi_set_gets_events() and set the
337"val" to non-zero. Any events that have been received by the driver
338since startup will immediately be delivered to the first user that
339registers for events. After that, if multiple users are registered
340for events, they will all receive all events that come in.
341
342For receiving commands, you have to individually register commands you
343want to receive. Call ipmi_register_for_cmd() and supply the netfn
Corey Minyardc69c3122006-09-30 23:27:56 -0700344and command name for each command you want to receive. You also
345specify a bitmask of the channels you want to receive the command from
346(or use IPMI_CHAN_ALL for all channels if you don't care). Only one
347user may be registered for each netfn/cmd/channel, but different users
348may register for different commands, or the same command if the
349channel bitmasks do not overlap.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350
351From userland, equivalent IOCTLs are provided to do these functions.
352
353
354The Lower Layer (SMI) Interface
355-------------------------------
356
357As mentioned before, multiple SMI interfaces may be registered to the
358message handler, each of these is assigned an interface number when
359they register with the message handler. They are generally assigned
360in the order they register, although if an SMI unregisters and then
361another one registers, all bets are off.
362
363The ipmi_smi.h defines the interface for management interfaces, see
364that for more details.
365
366
367The SI Driver
368-------------
369
Corey Minyardc11daf62016-10-25 20:31:53 -0500370The SI driver allows KCS, BT, and SMIC interfaces to be configured
371in the system. It discovers interfaces through a host of different
372methods, depending on the system.
373
374You can specify up to four interfaces on the module load line and
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300375control some module parameters::
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376
377 modprobe ipmi_si.o type=<type1>,<type2>....
378 ports=<port1>,<port2>... addrs=<addr1>,<addr2>...
Corey Minyardf2afae42013-02-27 17:05:13 -0800379 irqs=<irq1>,<irq2>...
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 regspacings=<sp1>,<sp2>,... regsizes=<size1>,<size2>,...
381 regshifts=<shift1>,<shift2>,...
382 slave_addrs=<addr1>,<addr2>,...
Corey Minyarda51f4a82006-10-03 01:13:59 -0700383 force_kipmid=<enable1>,<enable2>,...
Martin Wilckae74e822010-03-10 15:23:06 -0800384 kipmid_max_busy_us=<ustime1>,<ustime2>,...
Corey Minyardb361e272006-12-06 20:41:07 -0800385 unload_when_empty=[0|1]
Corey Minyardc11daf62016-10-25 20:31:53 -0500386 trydmi=[0|1] tryacpi=[0|1]
Corey Minyardf2afae42013-02-27 17:05:13 -0800387 tryplatform=[0|1] trypci=[0|1]
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388
Corey Minyardf2afae42013-02-27 17:05:13 -0800389Each of these except try... items is a list, the first item for the
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390first interface, second item for the second interface, etc.
391
392The si_type may be either "kcs", "smic", or "bt". If you leave it blank, it
393defaults to "kcs".
394
Corey Minyardf2afae42013-02-27 17:05:13 -0800395If you specify addrs as non-zero for an interface, the driver will
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396use the memory address given as the address of the device. This
397overrides si_ports.
398
Corey Minyardf2afae42013-02-27 17:05:13 -0800399If you specify ports as non-zero for an interface, the driver will
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400use the I/O port given as the device address.
401
Corey Minyardf2afae42013-02-27 17:05:13 -0800402If you specify irqs as non-zero for an interface, the driver will
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403attempt to use the given interrupt for the device.
404
Corey Minyardf2afae42013-02-27 17:05:13 -0800405The other try... items disable discovery by their corresponding
406names. These are all enabled by default, set them to zero to disable
407them. The tryplatform disables openfirmware.
408
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409The next three parameters have to do with register layout. The
410registers used by the interfaces may not appear at successive
411locations and they may not be in 8-bit registers. These parameters
412allow the layout of the data in the registers to be more precisely
413specified.
414
415The regspacings parameter give the number of bytes between successive
416register start addresses. For instance, if the regspacing is set to 4
417and the start address is 0xca2, then the address for the second
418register would be 0xca6. This defaults to 1.
419
420The regsizes parameter gives the size of a register, in bytes. The
421data used by IPMI is 8-bits wide, but it may be inside a larger
422register. This parameter allows the read and write type to specified.
423It may be 1, 2, 4, or 8. The default is 1.
424
425Since the register size may be larger than 32 bits, the IPMI data may not
426be in the lower 8 bits. The regshifts parameter give the amount to shift
427the data to get to the actual IPMI data.
428
429The slave_addrs specifies the IPMI address of the local BMC. This is
430usually 0x20 and the driver defaults to that, but in case it's not, it
431can be specified when the driver starts up.
432
Corey Minyarda51f4a82006-10-03 01:13:59 -0700433The force_ipmid parameter forcefully enables (if set to 1) or disables
434(if set to 0) the kernel IPMI daemon. Normally this is auto-detected
435by the driver, but systems with broken interrupts might need an enable,
436or users that don't want the daemon (don't need the performance, don't
437want the CPU hit) can disable it.
438
Corey Minyardb361e272006-12-06 20:41:07 -0800439If unload_when_empty is set to 1, the driver will be unloaded if it
440doesn't find any interfaces or all the interfaces fail to work. The
441default is one. Setting to 0 is useful with the hotmod, but is
442obviously only useful for modules.
443
Corey Minyarda51f4a82006-10-03 01:13:59 -0700444When compiled into the kernel, the parameters can be specified on the
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300445kernel command line as::
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446
447 ipmi_si.type=<type1>,<type2>...
448 ipmi_si.ports=<port1>,<port2>... ipmi_si.addrs=<addr1>,<addr2>...
Corey Minyardc11daf62016-10-25 20:31:53 -0500449 ipmi_si.irqs=<irq1>,<irq2>...
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 ipmi_si.regspacings=<sp1>,<sp2>,...
451 ipmi_si.regsizes=<size1>,<size2>,...
452 ipmi_si.regshifts=<shift1>,<shift2>,...
453 ipmi_si.slave_addrs=<addr1>,<addr2>,...
Corey Minyarda51f4a82006-10-03 01:13:59 -0700454 ipmi_si.force_kipmid=<enable1>,<enable2>,...
Martin Wilckae74e822010-03-10 15:23:06 -0800455 ipmi_si.kipmid_max_busy_us=<ustime1>,<ustime2>,...
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456
457It works the same as the module parameters of the same names.
458
Corey Minyard650dd0c2007-10-18 03:07:09 -0700459If your IPMI interface does not support interrupts and is a KCS or
460SMIC interface, the IPMI driver will start a kernel thread for the
461interface to help speed things up. This is a low-priority kernel
462thread that constantly polls the IPMI driver while an IPMI operation
463is in progress. The force_kipmid module parameter will all the user to
464force this thread on or off. If you force it off and don't have
465interrupts, the driver will run VERY slowly. Don't blame me,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466these interfaces suck.
467
Martin Wilckae74e822010-03-10 15:23:06 -0800468Unfortunately, this thread can use a lot of CPU depending on the
469interface's performance. This can waste a lot of CPU and cause
470various issues with detecting idle CPU and using extra power. To
471avoid this, the kipmid_max_busy_us sets the maximum amount of time, in
472microseconds, that kipmid will spin before sleeping for a tick. This
473value sets a balance between performance and CPU waste and needs to be
474tuned to your needs. Maybe, someday, auto-tuning will be added, but
475that's not a simple thing and even the auto-tuning would need to be
476tuned to the user's desired performance.
477
Corey Minyardb361e272006-12-06 20:41:07 -0800478The driver supports a hot add and remove of interfaces. This way,
479interfaces can be added or removed after the kernel is up and running.
Corey Minyard650dd0c2007-10-18 03:07:09 -0700480This is done using /sys/modules/ipmi_si/parameters/hotmod, which is a
481write-only parameter. You write a string to this interface. The string
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300482has the format::
483
Corey Minyardb361e272006-12-06 20:41:07 -0800484 <op1>[:op2[:op3...]]
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300485
486The "op"s are::
487
Corey Minyardb361e272006-12-06 20:41:07 -0800488 add|remove,kcs|bt|smic,mem|i/o,<address>[,<opt1>[,<opt2>[,...]]]
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300489
490You can specify more than one interface on the line. The "opt"s are::
491
Corey Minyardb361e272006-12-06 20:41:07 -0800492 rsp=<regspacing>
493 rsi=<regsize>
494 rsh=<regshift>
495 irq=<irq>
496 ipmb=<ipmb slave addr>
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300497
Corey Minyardb361e272006-12-06 20:41:07 -0800498and these have the same meanings as discussed above. Note that you
499can also use this on the kernel command line for a more compact format
500for specifying an interface. Note that when removing an interface,
501only the first three parameters (si type, address type, and address)
502are used for the comparison. Any options are ignored for removing.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503
Corey Minyard25930702012-03-19 16:00:55 -0500504The SMBus Driver (SSIF)
505-----------------------
506
507The SMBus driver allows up to 4 SMBus devices to be configured in the
508system. By default, the driver will only register with something it
509finds in DMI or ACPI tables. You can change this
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300510at module load time (for a module) with::
Corey Minyard25930702012-03-19 16:00:55 -0500511
512 modprobe ipmi_ssif.o
513 addr=<i2caddr1>[,<i2caddr2>[,...]]
514 adapter=<adapter1>[,<adapter2>[...]]
515 dbg=<flags1>,<flags2>...
Corey Minyardc11daf62016-10-25 20:31:53 -0500516 slave_addrs=<addr1>,<addr2>,...
517 tryacpi=[0|1] trydmi=[0|1]
Corey Minyard25930702012-03-19 16:00:55 -0500518 [dbg_probe=1]
519
520The addresses are normal I2C addresses. The adapter is the string
521name of the adapter, as shown in /sys/class/i2c-adapter/i2c-<n>/name.
Corey Minyardb0e9aaa2015-03-31 12:48:53 -0500522It is *NOT* i2c-<n> itself. Also, the comparison is done ignoring
523spaces, so if the name is "This is an I2C chip" you can say
524adapter_name=ThisisanI2cchip. This is because it's hard to pass in
525spaces in kernel parameters.
Corey Minyard25930702012-03-19 16:00:55 -0500526
527The debug flags are bit flags for each BMC found, they are:
528IPMI messages: 1, driver state: 2, timing: 4, I2C probe: 8
529
Corey Minyardc11daf62016-10-25 20:31:53 -0500530The tryxxx parameters can be used to disable detecting interfaces
531from various sources.
532
Corey Minyard25930702012-03-19 16:00:55 -0500533Setting dbg_probe to 1 will enable debugging of the probing and
534detection process for BMCs on the SMBusses.
535
536The slave_addrs specifies the IPMI address of the local BMC. This is
537usually 0x20 and the driver defaults to that, but in case it's not, it
538can be specified when the driver starts up.
539
540Discovering the IPMI compliant BMC on the SMBus can cause devices on
541the I2C bus to fail. The SMBus driver writes a "Get Device ID" IPMI
542message as a block write to the I2C bus and waits for a response.
543This action can be detrimental to some I2C devices. It is highly
544recommended that the known I2C address be given to the SMBus driver in
545the smb_addr parameter unless you have DMI or ACPI data to tell the
546driver what to use.
547
548When compiled into the kernel, the addresses can be specified on the
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300549kernel command line as::
Corey Minyard25930702012-03-19 16:00:55 -0500550
551 ipmb_ssif.addr=<i2caddr1>[,<i2caddr2>[...]]
552 ipmi_ssif.adapter=<adapter1>[,<adapter2>[...]]
553 ipmi_ssif.dbg=<flags1>[,<flags2>[...]]
554 ipmi_ssif.dbg_probe=1
Corey Minyardc11daf62016-10-25 20:31:53 -0500555 ipmi_ssif.slave_addrs=<addr1>[,<addr2>[...]]
556 ipmi_ssif.tryacpi=[0|1] ipmi_ssif.trydmi=[0|1]
Corey Minyard25930702012-03-19 16:00:55 -0500557
558These are the same options as on the module command line.
559
560The I2C driver does not support non-blocking access or polling, so
561this driver cannod to IPMI panic events, extend the watchdog at panic
562time, or other panic-related IPMI functions without special kernel
563patches and driver modifications. You can get those at the openipmi
564web page.
565
566The driver supports a hot add and remove of interfaces through the I2C
567sysfs interface.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568
569Other Pieces
570------------
571
Zhao Yakui37bf5012010-12-08 10:10:17 +0800572Get the detailed info related with the IPMI device
573--------------------------------------------------
574
575Some users need more detailed information about a device, like where
576the address came from or the raw base device for the IPMI interface.
577You can use the IPMI smi_watcher to catch the IPMI interfaces as they
578come or go, and to grab the information, you can use the function
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300579ipmi_get_smi_info(), which returns the following structure::
Zhao Yakui37bf5012010-12-08 10:10:17 +0800580
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300581 struct ipmi_smi_info {
Zhao Yakui37bf5012010-12-08 10:10:17 +0800582 enum ipmi_addr_src addr_src;
583 struct device *dev;
584 union {
585 struct {
586 void *acpi_handle;
587 } acpi_info;
588 } addr_info;
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300589 };
Zhao Yakui37bf5012010-12-08 10:10:17 +0800590
591Currently special info for only for SI_ACPI address sources is
592returned. Others may be added as necessary.
593
594Note that the dev pointer is included in the above structure, and
595assuming ipmi_smi_get_info returns success, you must call put_device
596on the dev pointer.
597
598
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599Watchdog
600--------
601
602A watchdog timer is provided that implements the Linux-standard
603watchdog timer interface. It has three module parameters that can be
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300604used to control it::
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605
606 modprobe ipmi_watchdog timeout=<t> pretimeout=<t> action=<action type>
607 preaction=<preaction type> preop=<preop type> start_now=x
Jean-Yves Fayec7f42c62015-09-29 11:39:19 +0200608 nowayout=x ifnum_to_use=n panic_wdt_timeout=<t>
Corey Minyardb2c03942006-12-06 20:41:00 -0800609
610ifnum_to_use specifies which interface the watchdog timer should use.
611The default is -1, which means to pick the first one registered.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612
613The timeout is the number of seconds to the action, and the pretimeout
614is the amount of seconds before the reset that the pre-timeout panic will
615occur (if pretimeout is zero, then pretimeout will not be enabled). Note
616that the pretimeout is the time before the final timeout. So if the
617timeout is 50 seconds and the pretimeout is 10 seconds, then the pretimeout
Jean-Yves Fayec7f42c62015-09-29 11:39:19 +0200618will occur in 40 second (10 seconds before the timeout). The panic_wdt_timeout
619is the value of timeout which is set on kernel panic, in order to let actions
620such as kdump to occur during panic.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621
622The action may be "reset", "power_cycle", or "power_off", and
623specifies what to do when the timer times out, and defaults to
624"reset".
625
626The preaction may be "pre_smi" for an indication through the SMI
627interface, "pre_int" for an indication through the SMI with an
628interrupts, and "pre_nmi" for a NMI on a preaction. This is how
629the driver is informed of the pretimeout.
630
631The preop may be set to "preop_none" for no operation on a pretimeout,
632"preop_panic" to set the preoperation to panic, or "preop_give_data"
633to provide data to read from the watchdog device when the pretimeout
634occurs. A "pre_nmi" setting CANNOT be used with "preop_give_data"
635because you can't do data operations from an NMI.
636
637When preop is set to "preop_give_data", one byte comes ready to read
638on the device when the pretimeout occurs. Select and fasync work on
639the device, as well.
640
641If start_now is set to 1, the watchdog timer will start running as
642soon as the driver is loaded.
643
644If nowayout is set to 1, the watchdog timer will not stop when the
645watchdog device is closed. The default value of nowayout is true
646if the CONFIG_WATCHDOG_NOWAYOUT option is enabled, or false if not.
647
648When compiled into the kernel, the kernel command line is available
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300649for configuring the watchdog::
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650
651 ipmi_watchdog.timeout=<t> ipmi_watchdog.pretimeout=<t>
652 ipmi_watchdog.action=<action type>
653 ipmi_watchdog.preaction=<preaction type>
654 ipmi_watchdog.preop=<preop type>
655 ipmi_watchdog.start_now=x
656 ipmi_watchdog.nowayout=x
Jean-Yves Fayec7f42c62015-09-29 11:39:19 +0200657 ipmi_watchdog.panic_wdt_timeout=<t>
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658
659The options are the same as the module parameter options.
660
661The watchdog will panic and start a 120 second reset timeout if it
662gets a pre-action. During a panic or a reboot, the watchdog will
663start a 120 timer if it is running to make sure the reboot occurs.
664
Corey Minyard612b5a82007-10-18 03:07:10 -0700665Note that if you use the NMI preaction for the watchdog, you MUST NOT
666use the nmi watchdog. There is no reasonable way to tell if an NMI
667comes from the IPMI controller, so it must assume that if it gets an
668otherwise unhandled NMI, it must be from IPMI and it will panic
669immediately.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700670
671Once you open the watchdog timer, you must write a 'V' character to the
672device to close it, or the timer will not stop. This is a new semantic
673for the driver, but makes it consistent with the rest of the watchdog
674drivers in Linux.
Corey Minyard845e78a2005-06-23 22:01:39 -0700675
676
677Panic Timeouts
678--------------
679
680The OpenIPMI driver supports the ability to put semi-custom and custom
681events in the system event log if a panic occurs. if you enable the
682'Generate a panic event to all BMCs on a panic' option, you will get
683one event on a panic in a standard IPMI event format. If you enable
684the 'Generate OEM events containing the panic string' option, you will
685also get a bunch of OEM events holding the panic string.
686
687
688The field settings of the events are:
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300689
Corey Minyard845e78a2005-06-23 22:01:39 -0700690* Generator ID: 0x21 (kernel)
691* EvM Rev: 0x03 (this event is formatting in IPMI 1.0 format)
692* Sensor Type: 0x20 (OS critical stop sensor)
693* Sensor #: The first byte of the panic string (0 if no panic string)
694* Event Dir | Event Type: 0x6f (Assertion, sensor-specific event info)
695* Event Data 1: 0xa1 (Runtime stop in OEM bytes 2 and 3)
696* Event data 2: second byte of panic string
697* Event data 3: third byte of panic string
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300698
Corey Minyard845e78a2005-06-23 22:01:39 -0700699See the IPMI spec for the details of the event layout. This event is
700always sent to the local management controller. It will handle routing
701the message to the right place
702
703Other OEM events have the following format:
Mauro Carvalho Chehabf5981a52017-05-14 15:18:08 -0300704
705* Record ID (bytes 0-1): Set by the SEL.
706* Record type (byte 2): 0xf0 (OEM non-timestamped)
707* byte 3: The slave address of the card saving the panic
708* byte 4: A sequence number (starting at zero)
709 The rest of the bytes (11 bytes) are the panic string. If the panic string
710 is longer than 11 bytes, multiple messages will be sent with increasing
711 sequence numbers.
Corey Minyard845e78a2005-06-23 22:01:39 -0700712
713Because you cannot send OEM events using the standard interface, this
714function will attempt to find an SEL and add the events there. It
715will first query the capabilities of the local management controller.
716If it has an SEL, then they will be stored in the SEL of the local
717management controller. If not, and the local management controller is
718an event generator, the event receiver from the local management
719controller will be queried and the events sent to the SEL on that
720device. Otherwise, the events go nowhere since there is nowhere to
721send them.
Corey Minyard3b625942005-06-23 22:01:42 -0700722
723
724Poweroff
725--------
726
727If the poweroff capability is selected, the IPMI driver will install
728a shutdown function into the standard poweroff function pointer. This
729is in the ipmi_poweroff module. When the system requests a powerdown,
730it will send the proper IPMI commands to do this. This is supported on
731several platforms.
732
Corey Minyard8c702e12005-09-06 15:18:46 -0700733There is a module parameter named "poweroff_powercycle" that may
734either be zero (do a power down) or non-zero (do a power cycle, power
735the system off, then power it on in a few seconds). Setting
736ipmi_poweroff.poweroff_control=x will do the same thing on the kernel
737command line. The parameter is also available via the proc filesystem
738in /proc/sys/dev/ipmi/poweroff_powercycle. Note that if the system
739does not support power cycling, it will always do the power off.
Corey Minyard3b625942005-06-23 22:01:42 -0700740
Corey Minyardb2c03942006-12-06 20:41:00 -0800741The "ifnum_to_use" parameter specifies which interface the poweroff
742code should use. The default is -1, which means to pick the first one
743registered.
744
Corey Minyard3b625942005-06-23 22:01:42 -0700745Note that if you have ACPI enabled, the system will prefer using ACPI to
746power off.