ia64/linux-2.6.18-xen.hg

view Documentation/stable_api_nonsense.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
1 The Linux Kernel Driver Interface
2 (all of your questions answered and then some)
4 Greg Kroah-Hartman <greg@kroah.com>
6 This is being written to try to explain why Linux does not have a binary
7 kernel interface, nor does it have a stable kernel interface. Please
8 realize that this article describes the _in kernel_ interfaces, not the
9 kernel to userspace interfaces. The kernel to userspace interface is
10 the one that application programs use, the syscall interface. That
11 interface is _very_ stable over time, and will not break. I have old
12 programs that were built on a pre 0.9something kernel that still work
13 just fine on the latest 2.6 kernel release. This interface is the one
14 that users and application programmers can count on being stable.
17 Executive Summary
18 -----------------
19 You think you want a stable kernel interface, but you really do not, and
20 you don't even know it. What you want is a stable running driver, and
21 you get that only if your driver is in the main kernel tree. You also
22 get lots of other good benefits if your driver is in the main kernel
23 tree, all of which has made Linux into such a strong, stable, and mature
24 operating system which is the reason you are using it in the first
25 place.
28 Intro
29 -----
31 It's only the odd person who wants to write a kernel driver that needs
32 to worry about the in-kernel interfaces changing. For the majority of
33 the world, they neither see this interface, nor do they care about it at
34 all.
36 First off, I'm not going to address _any_ legal issues about closed
37 source, hidden source, binary blobs, source wrappers, or any other term
38 that describes kernel drivers that do not have their source code
39 released under the GPL. Please consult a lawyer if you have any legal
40 questions, I'm a programmer and hence, I'm just going to be describing
41 the technical issues here (not to make light of the legal issues, they
42 are real, and you do need to be aware of them at all times.)
44 So, there are two main topics here, binary kernel interfaces and stable
45 kernel source interfaces. They both depend on each other, but we will
46 discuss the binary stuff first to get it out of the way.
49 Binary Kernel Interface
50 -----------------------
51 Assuming that we had a stable kernel source interface for the kernel, a
52 binary interface would naturally happen too, right? Wrong. Please
53 consider the following facts about the Linux kernel:
54 - Depending on the version of the C compiler you use, different kernel
55 data structures will contain different alignment of structures, and
56 possibly include different functions in different ways (putting
57 functions inline or not.) The individual function organization
58 isn't that important, but the different data structure padding is
59 very important.
60 - Depending on what kernel build options you select, a wide range of
61 different things can be assumed by the kernel:
62 - different structures can contain different fields
63 - Some functions may not be implemented at all, (i.e. some locks
64 compile away to nothing for non-SMP builds.)
65 - Parameter passing of variables from function to function can be
66 done in different ways (the CONFIG_REGPARM option controls
67 this.)
68 - Memory within the kernel can be aligned in different ways,
69 depending on the build options.
70 - Linux runs on a wide range of different processor architectures.
71 There is no way that binary drivers from one architecture will run
72 on another architecture properly.
74 Now a number of these issues can be addressed by simply compiling your
75 module for the exact specific kernel configuration, using the same exact
76 C compiler that the kernel was built with. This is sufficient if you
77 want to provide a module for a specific release version of a specific
78 Linux distribution. But multiply that single build by the number of
79 different Linux distributions and the number of different supported
80 releases of the Linux distribution and you quickly have a nightmare of
81 different build options on different releases. Also realize that each
82 Linux distribution release contains a number of different kernels, all
83 tuned to different hardware types (different processor types and
84 different options), so for even a single release you will need to create
85 multiple versions of your module.
87 Trust me, you will go insane over time if you try to support this kind
88 of release, I learned this the hard way a long time ago...
91 Stable Kernel Source Interfaces
92 -------------------------------
94 This is a much more "volatile" topic if you talk to people who try to
95 keep a Linux kernel driver that is not in the main kernel tree up to
96 date over time.
98 Linux kernel development is continuous and at a rapid pace, never
99 stopping to slow down. As such, the kernel developers find bugs in
100 current interfaces, or figure out a better way to do things. If they do
101 that, they then fix the current interfaces to work better. When they do
102 so, function names may change, structures may grow or shrink, and
103 function parameters may be reworked. If this happens, all of the
104 instances of where this interface is used within the kernel are fixed up
105 at the same time, ensuring that everything continues to work properly.
107 As a specific examples of this, the in-kernel USB interfaces have
108 undergone at least three different reworks over the lifetime of this
109 subsystem. These reworks were done to address a number of different
110 issues:
111 - A change from a synchronous model of data streams to an asynchronous
112 one. This reduced the complexity of a number of drivers and
113 increased the throughput of all USB drivers such that we are now
114 running almost all USB devices at their maximum speed possible.
115 - A change was made in the way data packets were allocated from the
116 USB core by USB drivers so that all drivers now needed to provide
117 more information to the USB core to fix a number of documented
118 deadlocks.
120 This is in stark contrast to a number of closed source operating systems
121 which have had to maintain their older USB interfaces over time. This
122 provides the ability for new developers to accidentally use the old
123 interfaces and do things in improper ways, causing the stability of the
124 operating system to suffer.
126 In both of these instances, all developers agreed that these were
127 important changes that needed to be made, and they were made, with
128 relatively little pain. If Linux had to ensure that it preserve a
129 stable source interface, a new interface would have been created, and
130 the older, broken one would have had to be maintained over time, leading
131 to extra work for the USB developers. Since all Linux USB developers do
132 their work on their own time, asking programmers to do extra work for no
133 gain, for free, is not a possibility.
135 Security issues are also very important for Linux. When a
136 security issue is found, it is fixed in a very short amount of time. A
137 number of times this has caused internal kernel interfaces to be
138 reworked to prevent the security problem from occurring. When this
139 happens, all drivers that use the interfaces were also fixed at the
140 same time, ensuring that the security problem was fixed and could not
141 come back at some future time accidentally. If the internal interfaces
142 were not allowed to change, fixing this kind of security problem and
143 insuring that it could not happen again would not be possible.
145 Kernel interfaces are cleaned up over time. If there is no one using a
146 current interface, it is deleted. This ensures that the kernel remains
147 as small as possible, and that all potential interfaces are tested as
148 well as they can be (unused interfaces are pretty much impossible to
149 test for validity.)
152 What to do
153 ----------
155 So, if you have a Linux kernel driver that is not in the main kernel
156 tree, what are you, a developer, supposed to do? Releasing a binary
157 driver for every different kernel version for every distribution is a
158 nightmare, and trying to keep up with an ever changing kernel interface
159 is also a rough job.
161 Simple, get your kernel driver into the main kernel tree (remember we
162 are talking about GPL released drivers here, if your code doesn't fall
163 under this category, good luck, you are on your own here, you leech
164 <insert link to leech comment from Andrew and Linus here>.) If your
165 driver is in the tree, and a kernel interface changes, it will be fixed
166 up by the person who did the kernel change in the first place. This
167 ensures that your driver is always buildable, and works over time, with
168 very little effort on your part.
170 The very good side effects of having your driver in the main kernel tree
171 are:
172 - The quality of the driver will rise as the maintenance costs (to the
173 original developer) will decrease.
174 - Other developers will add features to your driver.
175 - Other people will find and fix bugs in your driver.
176 - Other people will find tuning opportunities in your driver.
177 - Other people will update the driver for you when external interface
178 changes require it.
179 - The driver automatically gets shipped in all Linux distributions
180 without having to ask the distros to add it.
182 As Linux supports a larger number of different devices "out of the box"
183 than any other operating system, and it supports these devices on more
184 different processor architectures than any other operating system, this
185 proven type of development model must be doing something right :)
189 ------
191 Thanks to Randy Dunlap, Andrew Morton, David Brownell, Hanna Linder,
192 Robert Love, and Nishanth Aravamudan for their review and comments on
193 early drafts of this paper.