ia64/linux-2.6.18-xen.hg

view Documentation/pci-error-recovery.txt @ 897:329ea0ccb344

balloon: try harder to balloon up under memory pressure.

Currently if the balloon driver is unable to increase the guest's
reservation it assumes the failure was due to reaching its full
allocation, gives up on the ballooning operation and records the limit
it reached as the "hard limit". The driver will not try again until
the target is set again (even to the same value).

However it is possible that ballooning has in fact failed due to
memory pressure in the host and therefore it is desirable to keep
attempting to reach the target in case memory becomes available. The
most likely scenario is that some guests are ballooning down while
others are ballooning up and therefore there is temporary memory
pressure while things stabilise. You would not expect a well behaved
toolstack to ask a domain to balloon to more than its allocation nor
would you expect it to deliberately over-commit memory by setting
balloon targets which exceed the total host memory.

This patch drops the concept of a hard limit and causes the balloon
driver to retry increasing the reservation on a timer in the same
manner as when decreasing the reservation.

Also if we partially succeed in increasing the reservation
(i.e. receive less pages than we asked for) then we may as well keep
those pages rather than returning them to Xen.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
author Keir Fraser <keir.fraser@citrix.com>
date Fri Jun 05 14:01:20 2009 +0100 (2009-06-05)
parents 831230e53067
children
line source
2 PCI Error Recovery
3 ------------------
4 February 2, 2006
6 Current document maintainer:
7 Linas Vepstas <linas@austin.ibm.com>
10 Many PCI bus controllers are able to detect a variety of hardware
11 PCI errors on the bus, such as parity errors on the data and address
12 busses, as well as SERR and PERR errors. Some of the more advanced
13 chipsets are able to deal with these errors; these include PCI-E chipsets,
14 and the PCI-host bridges found on IBM Power4 and Power5-based pSeries
15 boxes. A typical action taken is to disconnect the affected device,
16 halting all I/O to it. The goal of a disconnection is to avoid system
17 corruption; for example, to halt system memory corruption due to DMA's
18 to "wild" addresses. Typically, a reconnection mechanism is also
19 offered, so that the affected PCI device(s) are reset and put back
20 into working condition. The reset phase requires coordination
21 between the affected device drivers and the PCI controller chip.
22 This document describes a generic API for notifying device drivers
23 of a bus disconnection, and then performing error recovery.
24 This API is currently implemented in the 2.6.16 and later kernels.
26 Reporting and recovery is performed in several steps. First, when
27 a PCI hardware error has resulted in a bus disconnect, that event
28 is reported as soon as possible to all affected device drivers,
29 including multiple instances of a device driver on multi-function
30 cards. This allows device drivers to avoid deadlocking in spinloops,
31 waiting for some i/o-space register to change, when it never will.
32 It also gives the drivers a chance to defer incoming I/O as
33 needed.
35 Next, recovery is performed in several stages. Most of the complexity
36 is forced by the need to handle multi-function devices, that is,
37 devices that have multiple device drivers associated with them.
38 In the first stage, each driver is allowed to indicate what type
39 of reset it desires, the choices being a simple re-enabling of I/O
40 or requesting a hard reset (a full electrical #RST of the PCI card).
41 If any driver requests a full reset, that is what will be done.
43 After a full reset and/or a re-enabling of I/O, all drivers are
44 again notified, so that they may then perform any device setup/config
45 that may be required. After these have all completed, a final
46 "resume normal operations" event is sent out.
48 The biggest reason for choosing a kernel-based implementation rather
49 than a user-space implementation was the need to deal with bus
50 disconnects of PCI devices attached to storage media, and, in particular,
51 disconnects from devices holding the root file system. If the root
52 file system is disconnected, a user-space mechanism would have to go
53 through a large number of contortions to complete recovery. Almost all
54 of the current Linux file systems are not tolerant of disconnection
55 from/reconnection to their underlying block device. By contrast,
56 bus errors are easy to manage in the device driver. Indeed, most
57 device drivers already handle very similar recovery procedures;
58 for example, the SCSI-generic layer already provides significant
59 mechanisms for dealing with SCSI bus errors and SCSI bus resets.
62 Detailed Design
63 ---------------
64 Design and implementation details below, based on a chain of
65 public email discussions with Ben Herrenschmidt, circa 5 April 2005.
67 The error recovery API support is exposed to the driver in the form of
68 a structure of function pointers pointed to by a new field in struct
69 pci_driver. A driver that fails to provide the structure is "non-aware",
70 and the actual recovery steps taken are platform dependent. The
71 arch/powerpc implementation will simulate a PCI hotplug remove/add.
73 This structure has the form:
74 struct pci_error_handlers
75 {
76 int (*error_detected)(struct pci_dev *dev, enum pci_channel_state);
77 int (*mmio_enabled)(struct pci_dev *dev);
78 int (*link_reset)(struct pci_dev *dev);
79 int (*slot_reset)(struct pci_dev *dev);
80 void (*resume)(struct pci_dev *dev);
81 };
83 The possible channel states are:
84 enum pci_channel_state {
85 pci_channel_io_normal, /* I/O channel is in normal state */
86 pci_channel_io_frozen, /* I/O to channel is blocked */
87 pci_channel_io_perm_failure, /* PCI card is dead */
88 };
90 Possible return values are:
91 enum pci_ers_result {
92 PCI_ERS_RESULT_NONE, /* no result/none/not supported in device driver */
93 PCI_ERS_RESULT_CAN_RECOVER, /* Device driver can recover without slot reset */
94 PCI_ERS_RESULT_NEED_RESET, /* Device driver wants slot to be reset. */
95 PCI_ERS_RESULT_DISCONNECT, /* Device has completely failed, is unrecoverable */
96 PCI_ERS_RESULT_RECOVERED, /* Device driver is fully recovered and operational */
97 };
99 A driver does not have to implement all of these callbacks; however,
100 if it implements any, it must implement error_detected(). If a callback
101 is not implemented, the corresponding feature is considered unsupported.
102 For example, if mmio_enabled() and resume() aren't there, then it
103 is assumed that the driver is not doing any direct recovery and requires
104 a reset. If link_reset() is not implemented, the card is assumed as
105 not care about link resets. Typically a driver will want to know about
106 a slot_reset().
108 The actual steps taken by a platform to recover from a PCI error
109 event will be platform-dependent, but will follow the general
110 sequence described below.
112 STEP 0: Error Event
113 -------------------
114 PCI bus error is detect by the PCI hardware. On powerpc, the slot
115 is isolated, in that all I/O is blocked: all reads return 0xffffffff,
116 all writes are ignored.
119 STEP 1: Notification
120 --------------------
121 Platform calls the error_detected() callback on every instance of
122 every driver affected by the error.
124 At this point, the device might not be accessible anymore, depending on
125 the platform (the slot will be isolated on powerpc). The driver may
126 already have "noticed" the error because of a failing I/O, but this
127 is the proper "synchronization point", that is, it gives the driver
128 a chance to cleanup, waiting for pending stuff (timers, whatever, etc...)
129 to complete; it can take semaphores, schedule, etc... everything but
130 touch the device. Within this function and after it returns, the driver
131 shouldn't do any new IOs. Called in task context. This is sort of a
132 "quiesce" point. See note about interrupts at the end of this doc.
134 All drivers participating in this system must implement this call.
135 The driver must return one of the following result codes:
136 - PCI_ERS_RESULT_CAN_RECOVER:
137 Driver returns this if it thinks it might be able to recover
138 the HW by just banging IOs or if it wants to be given
139 a chance to extract some diagnostic information (see
140 mmio_enable, below).
141 - PCI_ERS_RESULT_NEED_RESET:
142 Driver returns this if it can't recover without a hard
143 slot reset.
144 - PCI_ERS_RESULT_DISCONNECT:
145 Driver returns this if it doesn't want to recover at all.
147 The next step taken will depend on the result codes returned by the
148 drivers.
150 If all drivers on the segment/slot return PCI_ERS_RESULT_CAN_RECOVER,
151 then the platform should re-enable IOs on the slot (or do nothing in
152 particular, if the platform doesn't isolate slots), and recovery
153 proceeds to STEP 2 (MMIO Enable).
155 If any driver requested a slot reset (by returning PCI_ERS_RESULT_NEED_RESET),
156 then recovery proceeds to STEP 4 (Slot Reset).
158 If the platform is unable to recover the slot, the next step
159 is STEP 6 (Permanent Failure).
161 >>> The current powerpc implementation assumes that a device driver will
162 >>> *not* schedule or semaphore in this routine; the current powerpc
163 >>> implementation uses one kernel thread to notify all devices;
164 >>> thus, if one device sleeps/schedules, all devices are affected.
165 >>> Doing better requires complex multi-threaded logic in the error
166 >>> recovery implementation (e.g. waiting for all notification threads
167 >>> to "join" before proceeding with recovery.) This seems excessively
168 >>> complex and not worth implementing.
170 >>> The current powerpc implementation doesn't much care if the device
171 >>> attempts I/O at this point, or not. I/O's will fail, returning
172 >>> a value of 0xff on read, and writes will be dropped. If the device
173 >>> driver attempts more than 10K I/O's to a frozen adapter, it will
174 >>> assume that the device driver has gone into an infinite loop, and
175 >>> it will panic the the kernel. There doesn't seem to be any other
176 >>> way of stopping a device driver that insists on spinning on I/O.
178 STEP 2: MMIO Enabled
179 -------------------
180 The platform re-enables MMIO to the device (but typically not the
181 DMA), and then calls the mmio_enabled() callback on all affected
182 device drivers.
184 This is the "early recovery" call. IOs are allowed again, but DMA is
185 not (hrm... to be discussed, I prefer not), with some restrictions. This
186 is NOT a callback for the driver to start operations again, only to
187 peek/poke at the device, extract diagnostic information, if any, and
188 eventually do things like trigger a device local reset or some such,
189 but not restart operations. This is callback is made if all drivers on
190 a segment agree that they can try to recover and if no automatic link reset
191 was performed by the HW. If the platform can't just re-enable IOs without
192 a slot reset or a link reset, it wont call this callback, and instead
193 will have gone directly to STEP 3 (Link Reset) or STEP 4 (Slot Reset)
195 >>> The following is proposed; no platform implements this yet:
196 >>> Proposal: All I/O's should be done _synchronously_ from within
197 >>> this callback, errors triggered by them will be returned via
198 >>> the normal pci_check_whatever() API, no new error_detected()
199 >>> callback will be issued due to an error happening here. However,
200 >>> such an error might cause IOs to be re-blocked for the whole
201 >>> segment, and thus invalidate the recovery that other devices
202 >>> on the same segment might have done, forcing the whole segment
203 >>> into one of the next states, that is, link reset or slot reset.
205 The driver should return one of the following result codes:
206 - PCI_ERS_RESULT_RECOVERED
207 Driver returns this if it thinks the device is fully
208 functional and thinks it is ready to start
209 normal driver operations again. There is no
210 guarantee that the driver will actually be
211 allowed to proceed, as another driver on the
212 same segment might have failed and thus triggered a
213 slot reset on platforms that support it.
215 - PCI_ERS_RESULT_NEED_RESET
216 Driver returns this if it thinks the device is not
217 recoverable in it's current state and it needs a slot
218 reset to proceed.
220 - PCI_ERS_RESULT_DISCONNECT
221 Same as above. Total failure, no recovery even after
222 reset driver dead. (To be defined more precisely)
224 The next step taken depends on the results returned by the drivers.
225 If all drivers returned PCI_ERS_RESULT_RECOVERED, then the platform
226 proceeds to either STEP3 (Link Reset) or to STEP 5 (Resume Operations).
228 If any driver returned PCI_ERS_RESULT_NEED_RESET, then the platform
229 proceeds to STEP 4 (Slot Reset)
231 >>> The current powerpc implementation does not implement this callback.
234 STEP 3: Link Reset
235 ------------------
236 The platform resets the link, and then calls the link_reset() callback
237 on all affected device drivers. This is a PCI-Express specific state
238 and is done whenever a non-fatal error has been detected that can be
239 "solved" by resetting the link. This call informs the driver of the
240 reset and the driver should check to see if the device appears to be
241 in working condition.
243 The driver is not supposed to restart normal driver I/O operations
244 at this point. It should limit itself to "probing" the device to
245 check it's recoverability status. If all is right, then the platform
246 will call resume() once all drivers have ack'd link_reset().
248 Result codes:
249 (identical to STEP 3 (MMIO Enabled)
251 The platform then proceeds to either STEP 4 (Slot Reset) or STEP 5
252 (Resume Operations).
254 >>> The current powerpc implementation does not implement this callback.
257 STEP 4: Slot Reset
258 ------------------
259 The platform performs a soft or hard reset of the device, and then
260 calls the slot_reset() callback.
262 A soft reset consists of asserting the adapter #RST line and then
263 restoring the PCI BAR's and PCI configuration header to a state
264 that is equivalent to what it would be after a fresh system
265 power-on followed by power-on BIOS/system firmware initialization.
266 If the platform supports PCI hotplug, then the reset might be
267 performed by toggling the slot electrical power off/on.
269 It is important for the platform to restore the PCI config space
270 to the "fresh poweron" state, rather than the "last state". After
271 a slot reset, the device driver will almost always use its standard
272 device initialization routines, and an unusual config space setup
273 may result in hung devices, kernel panics, or silent data corruption.
275 This call gives drivers the chance to re-initialize the hardware
276 (re-download firmware, etc.). At this point, the driver may assume
277 that he card is in a fresh state and is fully functional. In
278 particular, interrupt generation should work normally.
280 Drivers should not yet restart normal I/O processing operations
281 at this point. If all device drivers report success on this
282 callback, the platform will call resume() to complete the sequence,
283 and let the driver restart normal I/O processing.
285 A driver can still return a critical failure for this function if
286 it can't get the device operational after reset. If the platform
287 previously tried a soft reset, it might now try a hard reset (power
288 cycle) and then call slot_reset() again. It the device still can't
289 be recovered, there is nothing more that can be done; the platform
290 will typically report a "permanent failure" in such a case. The
291 device will be considered "dead" in this case.
293 Drivers for multi-function cards will need to coordinate among
294 themselves as to which driver instance will perform any "one-shot"
295 or global device initialization. For example, the Symbios sym53cxx2
296 driver performs device init only from PCI function 0:
298 + if (PCI_FUNC(pdev->devfn) == 0)
299 + sym_reset_scsi_bus(np, 0);
301 Result codes:
302 - PCI_ERS_RESULT_DISCONNECT
303 Same as above.
305 Platform proceeds either to STEP 5 (Resume Operations) or STEP 6 (Permanent
306 Failure).
308 >>> The current powerpc implementation does not currently try a
309 >>> power-cycle reset if the driver returned PCI_ERS_RESULT_DISCONNECT.
310 >>> However, it probably should.
313 STEP 5: Resume Operations
314 -------------------------
315 The platform will call the resume() callback on all affected device
316 drivers if all drivers on the segment have returned
317 PCI_ERS_RESULT_RECOVERED from one of the 3 previous callbacks.
318 The goal of this callback is to tell the driver to restart activity,
319 that everything is back and running. This callback does not return
320 a result code.
322 At this point, if a new error happens, the platform will restart
323 a new error recovery sequence.
325 STEP 6: Permanent Failure
326 -------------------------
327 A "permanent failure" has occurred, and the platform cannot recover
328 the device. The platform will call error_detected() with a
329 pci_channel_state value of pci_channel_io_perm_failure.
331 The device driver should, at this point, assume the worst. It should
332 cancel all pending I/O, refuse all new I/O, returning -EIO to
333 higher layers. The device driver should then clean up all of its
334 memory and remove itself from kernel operations, much as it would
335 during system shutdown.
337 The platform will typically notify the system operator of the
338 permanent failure in some way. If the device is hotplug-capable,
339 the operator will probably want to remove and replace the device.
340 Note, however, not all failures are truly "permanent". Some are
341 caused by over-heating, some by a poorly seated card. Many
342 PCI error events are caused by software bugs, e.g. DMA's to
343 wild addresses or bogus split transactions due to programming
344 errors. See the discussion in powerpc/eeh-pci-error-recovery.txt
345 for additional detail on real-life experience of the causes of
346 software errors.
349 Conclusion; General Remarks
350 ---------------------------
351 The way those callbacks are called is platform policy. A platform with
352 no slot reset capability may want to just "ignore" drivers that can't
353 recover (disconnect them) and try to let other cards on the same segment
354 recover. Keep in mind that in most real life cases, though, there will
355 be only one driver per segment.
357 Now, a note about interrupts. If you get an interrupt and your
358 device is dead or has been isolated, there is a problem :)
359 The current policy is to turn this into a platform policy.
360 That is, the recovery API only requires that:
362 - There is no guarantee that interrupt delivery can proceed from any
363 device on the segment starting from the error detection and until the
364 resume callback is sent, at which point interrupts are expected to be
365 fully operational.
367 - There is no guarantee that interrupt delivery is stopped, that is,
368 a driver that gets an interrupt after detecting an error, or that detects
369 an error within the interrupt handler such that it prevents proper
370 ack'ing of the interrupt (and thus removal of the source) should just
371 return IRQ_NOTHANDLED. It's up to the platform to deal with that
372 condition, typically by masking the IRQ source during the duration of
373 the error handling. It is expected that the platform "knows" which
374 interrupts are routed to error-management capable slots and can deal
375 with temporarily disabling that IRQ number during error processing (this
376 isn't terribly complex). That means some IRQ latency for other devices
377 sharing the interrupt, but there is simply no other way. High end
378 platforms aren't supposed to share interrupts between many devices
379 anyway :)
381 >>> Implementation details for the powerpc platform are discussed in
382 >>> the file Documentation/powerpc/eeh-pci-error-recovery.txt
384 >>> As of this writing, there are six device drivers with patches
385 >>> implementing error recovery. Not all of these patches are in
386 >>> mainline yet. These may be used as "examples":
387 >>>
388 >>> drivers/scsi/ipr.c
389 >>> drivers/scsi/sym53cxx_2
390 >>> drivers/next/e100.c
391 >>> drivers/net/e1000
392 >>> drivers/net/ixgb
393 >>> drivers/net/s2io.c
395 The End
396 -------