view Documentation/spinlocks.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 UPDATE March 21 2005 Amit Gud <gud@eth.net>
3 Macros SPIN_LOCK_UNLOCKED and RW_LOCK_UNLOCKED are deprecated and will be
4 removed soon. So for any new code dynamic initialization should be used:
6 spinlock_t xxx_lock;
7 rwlock_t xxx_rw_lock;
9 static int __init xxx_init(void)
10 {
11 spin_lock_init(&xxx_lock);
12 rwlock_init(&xxx_rw_lock);
13 ...
14 }
16 module_init(xxx_init);
18 Reasons for deprecation
19 - it hurts automatic lock validators
20 - it becomes intrusive for the realtime preemption patches
22 Following discussion is still valid, however, with the dynamic initialization
23 of spinlocks instead of static.
25 -----------------------
27 On Fri, 2 Jan 1998, Doug Ledford wrote:
28 >
29 > I'm working on making the aic7xxx driver more SMP friendly (as well as
30 > importing the latest FreeBSD sequencer code to have 7895 support) and wanted
31 > to get some info from you. The goal here is to make the various routines
32 > SMP safe as well as UP safe during interrupts and other manipulating
33 > routines. So far, I've added a spin_lock variable to things like my queue
34 > structs. Now, from what I recall, there are some spin lock functions I can
35 > use to lock these spin locks from other use as opposed to a (nasty)
36 > save_flags(); cli(); stuff; restore_flags(); construct. Where do I find
37 > these routines and go about making use of them? Do they only lock on a
38 > per-processor basis or can they also lock say an interrupt routine from
39 > mucking with a queue if the queue routine was manipulating it when the
40 > interrupt occurred, or should I still use a cli(); based construct on that
41 > one?
43 See <asm/spinlock.h>. The basic version is:
45 spinlock_t xxx_lock = SPIN_LOCK_UNLOCKED;
48 unsigned long flags;
50 spin_lock_irqsave(&xxx_lock, flags);
51 ... critical section here ..
52 spin_unlock_irqrestore(&xxx_lock, flags);
54 and the above is always safe. It will disable interrupts _locally_, but the
55 spinlock itself will guarantee the global lock, so it will guarantee that
56 there is only one thread-of-control within the region(s) protected by that
57 lock.
59 Note that it works well even under UP - the above sequence under UP
60 essentially is just the same as doing a
62 unsigned long flags;
64 save_flags(flags); cli();
65 ... critical section ...
66 restore_flags(flags);
68 so the code does _not_ need to worry about UP vs SMP issues: the spinlocks
69 work correctly under both (and spinlocks are actually more efficient on
70 architectures that allow doing the "save_flags + cli" in one go because I
71 don't export that interface normally).
73 NOTE NOTE NOTE! The reason the spinlock is so much faster than a global
74 interrupt lock under SMP is exactly because it disables interrupts only on
75 the local CPU. The spin-lock is safe only when you _also_ use the lock
76 itself to do locking across CPU's, which implies that EVERYTHING that
77 touches a shared variable has to agree about the spinlock they want to
78 use.
80 The above is usually pretty simple (you usually need and want only one
81 spinlock for most things - using more than one spinlock can make things a
82 lot more complex and even slower and is usually worth it only for
83 sequences that you _know_ need to be split up: avoid it at all cost if you
84 aren't sure). HOWEVER, it _does_ mean that if you have some code that does
86 cli();
87 .. critical section ..
88 sti();
90 and another sequence that does
92 spin_lock_irqsave(flags);
93 .. critical section ..
94 spin_unlock_irqrestore(flags);
96 then they are NOT mutually exclusive, and the critical regions can happen
97 at the same time on two different CPU's. That's fine per se, but the
98 critical regions had better be critical for different things (ie they
99 can't stomp on each other).
101 The above is a problem mainly if you end up mixing code - for example the
102 routines in ll_rw_block() tend to use cli/sti to protect the atomicity of
103 their actions, and if a driver uses spinlocks instead then you should
104 think about issues like the above..
106 This is really the only really hard part about spinlocks: once you start
107 using spinlocks they tend to expand to areas you might not have noticed
108 before, because you have to make sure the spinlocks correctly protect the
109 shared data structures _everywhere_ they are used. The spinlocks are most
110 easily added to places that are completely independent of other code (ie
111 internal driver data structures that nobody else ever touches, for
112 example).
114 ----
116 Lesson 2: reader-writer spinlocks.
118 If your data accesses have a very natural pattern where you usually tend
119 to mostly read from the shared variables, the reader-writer locks
120 (rw_lock) versions of the spinlocks are often nicer. They allow multiple
121 readers to be in the same critical region at once, but if somebody wants
122 to change the variables it has to get an exclusive write lock. The
123 routines look the same as above:
125 rwlock_t xxx_lock = RW_LOCK_UNLOCKED;
128 unsigned long flags;
130 read_lock_irqsave(&xxx_lock, flags);
131 .. critical section that only reads the info ...
132 read_unlock_irqrestore(&xxx_lock, flags);
134 write_lock_irqsave(&xxx_lock, flags);
135 .. read and write exclusive access to the info ...
136 write_unlock_irqrestore(&xxx_lock, flags);
138 The above kind of lock is useful for complex data structures like linked
139 lists etc, especially when you know that most of the work is to just
140 traverse the list searching for entries without changing the list itself,
141 for example. Then you can use the read lock for that kind of list
142 traversal, which allows many concurrent readers. Anything that _changes_
143 the list will have to get the write lock.
145 Note: you cannot "upgrade" a read-lock to a write-lock, so if you at _any_
146 time need to do any changes (even if you don't do it every time), you have
147 to get the write-lock at the very beginning. I could fairly easily add a
148 primitive to create a "upgradeable" read-lock, but it hasn't been an issue
149 yet. Tell me if you'd want one.
151 ----
153 Lesson 3: spinlocks revisited.
155 The single spin-lock primitives above are by no means the only ones. They
156 are the most safe ones, and the ones that work under all circumstances,
157 but partly _because_ they are safe they are also fairly slow. They are
158 much faster than a generic global cli/sti pair, but slower than they'd
159 need to be, because they do have to disable interrupts (which is just a
160 single instruction on a x86, but it's an expensive one - and on other
161 architectures it can be worse).
163 If you have a case where you have to protect a data structure across
164 several CPU's and you want to use spinlocks you can potentially use
165 cheaper versions of the spinlocks. IFF you know that the spinlocks are
166 never used in interrupt handlers, you can use the non-irq versions:
168 spin_lock(&lock);
169 ...
170 spin_unlock(&lock);
172 (and the equivalent read-write versions too, of course). The spinlock will
173 guarantee the same kind of exclusive access, and it will be much faster.
174 This is useful if you know that the data in question is only ever
175 manipulated from a "process context", ie no interrupts involved.
177 The reasons you mustn't use these versions if you have interrupts that
178 play with the spinlock is that you can get deadlocks:
180 spin_lock(&lock);
181 ...
182 <- interrupt comes in:
183 spin_lock(&lock);
185 where an interrupt tries to lock an already locked variable. This is ok if
186 the other interrupt happens on another CPU, but it is _not_ ok if the
187 interrupt happens on the same CPU that already holds the lock, because the
188 lock will obviously never be released (because the interrupt is waiting
189 for the lock, and the lock-holder is interrupted by the interrupt and will
190 not continue until the interrupt has been processed).
192 (This is also the reason why the irq-versions of the spinlocks only need
193 to disable the _local_ interrupts - it's ok to use spinlocks in interrupts
194 on other CPU's, because an interrupt on another CPU doesn't interrupt the
195 CPU that holds the lock, so the lock-holder can continue and eventually
196 releases the lock).
198 Note that you can be clever with read-write locks and interrupts. For
199 example, if you know that the interrupt only ever gets a read-lock, then
200 you can use a non-irq version of read locks everywhere - because they
201 don't block on each other (and thus there is no dead-lock wrt interrupts.
202 But when you do the write-lock, you have to use the irq-safe version.
204 For an example of being clever with rw-locks, see the "waitqueue_lock"
205 handling in kernel/sched.c - nothing ever _changes_ a wait-queue from
206 within an interrupt, they only read the queue in order to know whom to
207 wake up. So read-locks are safe (which is good: they are very common
208 indeed), while write-locks need to protect themselves against interrupts.
210 Linus