view Documentation/pcmcia/driver-changes.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
line source
1 This file details changes in 2.6 which affect PCMCIA card driver authors:
3 * New release helper (as of 2.6.17)
4 Instead of calling pcmcia_release_{configuration,io,irq,win}, all that's
5 necessary now is calling pcmcia_disable_device. As there is no valid
6 reason left to call pcmcia_release_io and pcmcia_release_irq, the
7 exports for them were removed.
9 * Unify detach and REMOVAL event code, as well as attach and INSERTION
10 code (as of 2.6.16)
11 void (*remove) (struct pcmcia_device *dev);
12 int (*probe) (struct pcmcia_device *dev);
14 * Move suspend, resume and reset out of event handler (as of 2.6.16)
15 int (*suspend) (struct pcmcia_device *dev);
16 int (*resume) (struct pcmcia_device *dev);
17 should be initialized in struct pcmcia_driver, and handle
20 * event handler initialization in struct pcmcia_driver (as of 2.6.13)
21 The event handler is notified of all events, and must be initialized
22 as the event() callback in the driver's struct pcmcia_driver.
24 * pcmcia/version.h should not be used (as of 2.6.13)
25 This file will be removed eventually.
27 * in-kernel device<->driver matching (as of 2.6.13)
28 PCMCIA devices and their correct drivers can now be matched in
29 kernelspace. See 'devicetable.txt' for details.
31 * Device model integration (as of 2.6.11)
32 A struct pcmcia_device is registered with the device model core,
33 and can be used (e.g. for SET_NETDEV_DEV) by using
34 handle_to_dev(client_handle_t * handle).
36 * Convert internal I/O port addresses to unsigned long (as of 2.6.11)
37 ioaddr_t should be replaced by kio_addr_t in PCMCIA card drivers.
39 * irq_mask and irq_list parameters (as of 2.6.11)
40 The irq_mask and irq_list parameters should no longer be used in
41 PCMCIA card drivers. Instead, it is the job of the PCMCIA core to
42 determine which IRQ should be used. Therefore, link->irq.IRQInfo2
43 is ignored.
45 * client->PendingEvents is gone (as of 2.6.11)
46 client->PendingEvents is no longer available.
48 * client->Attributes are gone (as of 2.6.11)
49 client->Attributes is unused, therefore it is removed from all
50 PCMCIA card drivers
52 * core functions no longer available (as of 2.6.11)
53 The following functions have been removed from the kernel source
54 because they are unused by all in-kernel drivers, and no external
55 driver was reported to rely on them:
56 pcmcia_get_first_region()
57 pcmcia_get_next_region()
58 pcmcia_modify_window()
59 pcmcia_set_event_mask()
60 pcmcia_get_first_window()
61 pcmcia_get_next_window()
63 * device list iteration upon module removal (as of 2.6.10)
64 It is no longer necessary to iterate on the driver's internal
65 client list and call the ->detach() function upon module removal.
67 * Resource management. (as of 2.6.8)
68 Although the PCMCIA subsystem will allocate resources for cards,
69 it no longer marks these resources busy. This means that driver
70 authors are now responsible for claiming your resources as per
71 other drivers in Linux. You should use request_region() to mark
72 your IO regions in-use, and request_mem_region() to mark your
73 memory regions in-use. The name argument should be a pointer to
74 your driver name. Eg, for pcnet_cs, name should point to the
75 string "pcnet_cs".
77 * CardServices is gone
78 CardServices() in 2.4 is just a big switch statement to call various
79 services. In 2.6, all of those entry points are exported and called
80 directly (except for pcmcia_report_error(), just use cs_error() instead).
82 * struct pcmcia_driver
83 You need to use struct pcmcia_driver and pcmcia_{un,}register_driver
84 instead of {un,}register_pccard_driver