Version:  2.0.40 2.2.26 2.4.37 3.13 3.14 3.15 3.16 3.17 3.18 3.19 4.0 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 4.10


  1 The padata parallel execution mechanism
  2 Last updated for 2.6.36
  4 Padata is a mechanism by which the kernel can farm work out to be done in
  5 parallel on multiple CPUs while retaining the ordering of tasks.  It was
  6 developed for use with the IPsec code, which needs to be able to perform
  7 encryption and decryption on large numbers of packets without reordering
  8 those packets.  The crypto developers made a point of writing padata in a
  9 sufficiently general fashion that it could be put to other uses as well.
 11 The first step in using padata is to set up a padata_instance structure for
 12 overall control of how tasks are to be run:
 14     #include <linux/padata.h>
 16     struct padata_instance *padata_alloc(struct workqueue_struct *wq,
 17                                          const struct cpumask *pcpumask,
 18                                          const struct cpumask *cbcpumask);
 20 The pcpumask describes which processors will be used to execute work
 21 submitted to this instance in parallel. The cbcpumask defines which
 22 processors are allowed to be used as the serialization callback processor.
 23 The workqueue wq is where the work will actually be done; it should be
 24 a multithreaded queue, naturally.
 26 To allocate a padata instance with the cpu_possible_mask for both
 27 cpumasks this helper function can be used:
 29     struct padata_instance *padata_alloc_possible(struct workqueue_struct *wq);
 31 Note: Padata maintains two kinds of cpumasks internally. The user supplied
 32 cpumasks, submitted by padata_alloc/padata_alloc_possible and the 'usable'
 33 cpumasks. The usable cpumasks are always a subset of active CPUs in the
 34 user supplied cpumasks; these are the cpumasks padata actually uses. So
 35 it is legal to supply a cpumask to padata that contains offline CPUs.
 36 Once an offline CPU in the user supplied cpumask comes online, padata
 37 is going to use it.
 39 There are functions for enabling and disabling the instance:
 41     int padata_start(struct padata_instance *pinst);
 42     void padata_stop(struct padata_instance *pinst);
 44 These functions are setting or clearing the "PADATA_INIT" flag;
 45 if that flag is not set, other functions will refuse to work.
 46 padata_start returns zero on success (flag set) or -EINVAL if the
 47 padata cpumask contains no active CPU (flag not set).
 48 padata_stop clears the flag and blocks until the padata instance
 49 is unused.
 51 The list of CPUs to be used can be adjusted with these functions:
 53     int padata_set_cpumasks(struct padata_instance *pinst,
 54                             cpumask_var_t pcpumask,
 55                             cpumask_var_t cbcpumask);
 56     int padata_set_cpumask(struct padata_instance *pinst, int cpumask_type,
 57                            cpumask_var_t cpumask);
 58     int padata_add_cpu(struct padata_instance *pinst, int cpu, int mask);
 59     int padata_remove_cpu(struct padata_instance *pinst, int cpu, int mask);
 61 Changing the CPU masks are expensive operations, though, so it should not be
 62 done with great frequency.
 64 It's possible to change both cpumasks of a padata instance with
 65 padata_set_cpumasks by specifying the cpumasks for parallel execution (pcpumask)
 66 and for the serial callback function (cbcpumask). padata_set_cpumask is used to
 67 change just one of the cpumasks. Here cpumask_type is one of PADATA_CPU_SERIAL,
 68 PADATA_CPU_PARALLEL and cpumask specifies the new cpumask to use.
 69 To simply add or remove one CPU from a certain cpumask the functions
 70 padata_add_cpu/padata_remove_cpu are used. cpu specifies the CPU to add or
 71 remove and mask is one of PADATA_CPU_SERIAL, PADATA_CPU_PARALLEL.
 73 If a user is interested in padata cpumask changes, he can register to
 74 the padata cpumask change notifier:
 76     int padata_register_cpumask_notifier(struct padata_instance *pinst,
 77                                          struct notifier_block *nblock);
 79 To unregister from that notifier:
 81     int padata_unregister_cpumask_notifier(struct padata_instance *pinst,
 82                                            struct notifier_block *nblock);
 84 The padata cpumask change notifier notifies about changes of the usable
 85 cpumasks, i.e. the subset of active CPUs in the user supplied cpumask.
 87 Padata calls the notifier chain with:
 89     blocking_notifier_call_chain(&pinst->cpumask_change_notifier,
 90                                  notification_mask,
 91                                  &pd_new->cpumask);
 93 Here cpumask_change_notifier is registered notifier, notification_mask
 94 is one of PADATA_CPU_SERIAL, PADATA_CPU_PARALLEL and cpumask is a pointer
 95 to a struct padata_cpumask that contains the new cpumask information.
 97 Actually submitting work to the padata instance requires the creation of a
 98 padata_priv structure:
100     struct padata_priv {
101         /* Other stuff here... */
102         void                    (*parallel)(struct padata_priv *padata);
103         void                    (*serial)(struct padata_priv *padata);
104     };
106 This structure will almost certainly be embedded within some larger
107 structure specific to the work to be done.  Most of its fields are private to
108 padata, but the structure should be zeroed at initialisation time, and the
109 parallel() and serial() functions should be provided.  Those functions will
110 be called in the process of getting the work done as we will see
111 momentarily.
113 The submission of work is done with:
115     int padata_do_parallel(struct padata_instance *pinst,
116                            struct padata_priv *padata, int cb_cpu);
118 The pinst and padata structures must be set up as described above; cb_cpu
119 specifies which CPU will be used for the final callback when the work is
120 done; it must be in the current instance's CPU mask.  The return value from
121 padata_do_parallel() is zero on success, indicating that the work is in
122 progress. -EBUSY means that somebody, somewhere else is messing with the
123 instance's CPU mask, while -EINVAL is a complaint about cb_cpu not being
124 in that CPU mask or about a not running instance.
126 Each task submitted to padata_do_parallel() will, in turn, be passed to
127 exactly one call to the above-mentioned parallel() function, on one CPU, so
128 true parallelism is achieved by submitting multiple tasks.  Despite the
129 fact that the workqueue is used to make these calls, parallel() is run with
130 software interrupts disabled and thus cannot sleep.  The parallel()
131 function gets the padata_priv structure pointer as its lone parameter;
132 information about the actual work to be done is probably obtained by using
133 container_of() to find the enclosing structure.
135 Note that parallel() has no return value; the padata subsystem assumes that
136 parallel() will take responsibility for the task from this point.  The work
137 need not be completed during this call, but, if parallel() leaves work
138 outstanding, it should be prepared to be called again with a new job before
139 the previous one completes.  When a task does complete, parallel() (or
140 whatever function actually finishes the job) should inform padata of the
141 fact with a call to:
143     void padata_do_serial(struct padata_priv *padata);
145 At some point in the future, padata_do_serial() will trigger a call to the
146 serial() function in the padata_priv structure.  That call will happen on
147 the CPU requested in the initial call to padata_do_parallel(); it, too, is
148 done through the workqueue, but with local software interrupts disabled.
149 Note that this call may be deferred for a while since the padata code takes
150 pains to ensure that tasks are completed in the order in which they were
151 submitted.
153 The one remaining function in the padata API should be called to clean up
154 when a padata instance is no longer needed:
156     void padata_free(struct padata_instance *pinst);
158 This function will busy-wait while any remaining tasks are completed, so it
159 might be best not to call it while there is work outstanding.  Shutting
160 down the workqueue, if necessary, should be done separately.

This page was automatically generated by LXR 0.3.1 (source).  •  Linux is a registered trademark of Linus Torvalds  •  Contact us