Commit 089545f0c71bab6511395c2a060d7f81a99bad58
Committed by
Linus Torvalds
1 parent
088c4ec16a
Exists in
master
and in
7 other branches
[PATCH] s390: cputime_t fixes
There are some more places where the use of cputime_t instead of an integer type and the associated macros is necessary for the virtual cputime accounting on s390. Affected are the s390 specific appldata code and BSD process accounting. Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Showing 2 changed files with 16 additions and 14 deletions Inline Diff
arch/s390/appldata/appldata_os.c
1 | /* | 1 | /* |
2 | * arch/s390/appldata/appldata_os.c | 2 | * arch/s390/appldata/appldata_os.c |
3 | * | 3 | * |
4 | * Data gathering module for Linux-VM Monitor Stream, Stage 1. | 4 | * Data gathering module for Linux-VM Monitor Stream, Stage 1. |
5 | * Collects misc. OS related data (CPU utilization, running processes). | 5 | * Collects misc. OS related data (CPU utilization, running processes). |
6 | * | 6 | * |
7 | * Copyright (C) 2003 IBM Corporation, IBM Deutschland Entwicklung GmbH. | 7 | * Copyright (C) 2003 IBM Corporation, IBM Deutschland Entwicklung GmbH. |
8 | * | 8 | * |
9 | * Author: Gerald Schaefer <geraldsc@de.ibm.com> | 9 | * Author: Gerald Schaefer <geraldsc@de.ibm.com> |
10 | */ | 10 | */ |
11 | 11 | ||
12 | #include <linux/config.h> | 12 | #include <linux/config.h> |
13 | #include <linux/module.h> | 13 | #include <linux/module.h> |
14 | #include <linux/init.h> | 14 | #include <linux/init.h> |
15 | #include <linux/slab.h> | 15 | #include <linux/slab.h> |
16 | #include <linux/errno.h> | 16 | #include <linux/errno.h> |
17 | #include <linux/kernel_stat.h> | 17 | #include <linux/kernel_stat.h> |
18 | #include <linux/netdevice.h> | 18 | #include <linux/netdevice.h> |
19 | #include <linux/sched.h> | 19 | #include <linux/sched.h> |
20 | #include <asm/smp.h> | 20 | #include <asm/smp.h> |
21 | 21 | ||
22 | #include "appldata.h" | 22 | #include "appldata.h" |
23 | 23 | ||
24 | 24 | ||
25 | #define MY_PRINT_NAME "appldata_os" /* for debug messages, etc. */ | 25 | #define MY_PRINT_NAME "appldata_os" /* for debug messages, etc. */ |
26 | #define LOAD_INT(x) ((x) >> FSHIFT) | 26 | #define LOAD_INT(x) ((x) >> FSHIFT) |
27 | #define LOAD_FRAC(x) LOAD_INT(((x) & (FIXED_1-1)) * 100) | 27 | #define LOAD_FRAC(x) LOAD_INT(((x) & (FIXED_1-1)) * 100) |
28 | 28 | ||
29 | /* | 29 | /* |
30 | * OS data | 30 | * OS data |
31 | * | 31 | * |
32 | * This is accessed as binary data by z/VM. If changes to it can't be avoided, | 32 | * This is accessed as binary data by z/VM. If changes to it can't be avoided, |
33 | * the structure version (product ID, see appldata_base.c) needs to be changed | 33 | * the structure version (product ID, see appldata_base.c) needs to be changed |
34 | * as well and all documentation and z/VM applications using it must be | 34 | * as well and all documentation and z/VM applications using it must be |
35 | * updated. | 35 | * updated. |
36 | * | 36 | * |
37 | * The record layout is documented in the Linux for zSeries Device Drivers | 37 | * The record layout is documented in the Linux for zSeries Device Drivers |
38 | * book: | 38 | * book: |
39 | * http://oss.software.ibm.com/developerworks/opensource/linux390/index.shtml | 39 | * http://oss.software.ibm.com/developerworks/opensource/linux390/index.shtml |
40 | */ | 40 | */ |
41 | struct appldata_os_per_cpu { | 41 | struct appldata_os_per_cpu { |
42 | u32 per_cpu_user; /* timer ticks spent in user mode */ | 42 | u32 per_cpu_user; /* timer ticks spent in user mode */ |
43 | u32 per_cpu_nice; /* ... spent with modified priority */ | 43 | u32 per_cpu_nice; /* ... spent with modified priority */ |
44 | u32 per_cpu_system; /* ... spent in kernel mode */ | 44 | u32 per_cpu_system; /* ... spent in kernel mode */ |
45 | u32 per_cpu_idle; /* ... spent in idle mode */ | 45 | u32 per_cpu_idle; /* ... spent in idle mode */ |
46 | 46 | ||
47 | // New in 2.6 --> | 47 | // New in 2.6 --> |
48 | u32 per_cpu_irq; /* ... spent in interrupts */ | 48 | u32 per_cpu_irq; /* ... spent in interrupts */ |
49 | u32 per_cpu_softirq; /* ... spent in softirqs */ | 49 | u32 per_cpu_softirq; /* ... spent in softirqs */ |
50 | u32 per_cpu_iowait; /* ... spent while waiting for I/O */ | 50 | u32 per_cpu_iowait; /* ... spent while waiting for I/O */ |
51 | // <-- New in 2.6 | 51 | // <-- New in 2.6 |
52 | } __attribute__((packed)); | 52 | } __attribute__((packed)); |
53 | 53 | ||
54 | struct appldata_os_data { | 54 | struct appldata_os_data { |
55 | u64 timestamp; | 55 | u64 timestamp; |
56 | u32 sync_count_1; /* after VM collected the record data, */ | 56 | u32 sync_count_1; /* after VM collected the record data, */ |
57 | u32 sync_count_2; /* sync_count_1 and sync_count_2 should be the | 57 | u32 sync_count_2; /* sync_count_1 and sync_count_2 should be the |
58 | same. If not, the record has been updated on | 58 | same. If not, the record has been updated on |
59 | the Linux side while VM was collecting the | 59 | the Linux side while VM was collecting the |
60 | (possibly corrupt) data */ | 60 | (possibly corrupt) data */ |
61 | 61 | ||
62 | u32 nr_cpus; /* number of (virtual) CPUs */ | 62 | u32 nr_cpus; /* number of (virtual) CPUs */ |
63 | u32 per_cpu_size; /* size of the per-cpu data struct */ | 63 | u32 per_cpu_size; /* size of the per-cpu data struct */ |
64 | u32 cpu_offset; /* offset of the first per-cpu data struct */ | 64 | u32 cpu_offset; /* offset of the first per-cpu data struct */ |
65 | 65 | ||
66 | u32 nr_running; /* number of runnable threads */ | 66 | u32 nr_running; /* number of runnable threads */ |
67 | u32 nr_threads; /* number of threads */ | 67 | u32 nr_threads; /* number of threads */ |
68 | u32 avenrun[3]; /* average nr. of running processes during */ | 68 | u32 avenrun[3]; /* average nr. of running processes during */ |
69 | /* the last 1, 5 and 15 minutes */ | 69 | /* the last 1, 5 and 15 minutes */ |
70 | 70 | ||
71 | // New in 2.6 --> | 71 | // New in 2.6 --> |
72 | u32 nr_iowait; /* number of blocked threads | 72 | u32 nr_iowait; /* number of blocked threads |
73 | (waiting for I/O) */ | 73 | (waiting for I/O) */ |
74 | // <-- New in 2.6 | 74 | // <-- New in 2.6 |
75 | 75 | ||
76 | /* per cpu data */ | 76 | /* per cpu data */ |
77 | struct appldata_os_per_cpu os_cpu[0]; | 77 | struct appldata_os_per_cpu os_cpu[0]; |
78 | } __attribute__((packed)); | 78 | } __attribute__((packed)); |
79 | 79 | ||
80 | static struct appldata_os_data *appldata_os_data; | 80 | static struct appldata_os_data *appldata_os_data; |
81 | 81 | ||
82 | 82 | ||
83 | static inline void appldata_print_debug(struct appldata_os_data *os_data) | 83 | static inline void appldata_print_debug(struct appldata_os_data *os_data) |
84 | { | 84 | { |
85 | int a0, a1, a2, i; | 85 | int a0, a1, a2, i; |
86 | 86 | ||
87 | P_DEBUG("--- OS - RECORD ---\n"); | 87 | P_DEBUG("--- OS - RECORD ---\n"); |
88 | P_DEBUG("nr_threads = %u\n", os_data->nr_threads); | 88 | P_DEBUG("nr_threads = %u\n", os_data->nr_threads); |
89 | P_DEBUG("nr_running = %u\n", os_data->nr_running); | 89 | P_DEBUG("nr_running = %u\n", os_data->nr_running); |
90 | P_DEBUG("nr_iowait = %u\n", os_data->nr_iowait); | 90 | P_DEBUG("nr_iowait = %u\n", os_data->nr_iowait); |
91 | P_DEBUG("avenrun(int) = %8x / %8x / %8x\n", os_data->avenrun[0], | 91 | P_DEBUG("avenrun(int) = %8x / %8x / %8x\n", os_data->avenrun[0], |
92 | os_data->avenrun[1], os_data->avenrun[2]); | 92 | os_data->avenrun[1], os_data->avenrun[2]); |
93 | a0 = os_data->avenrun[0]; | 93 | a0 = os_data->avenrun[0]; |
94 | a1 = os_data->avenrun[1]; | 94 | a1 = os_data->avenrun[1]; |
95 | a2 = os_data->avenrun[2]; | 95 | a2 = os_data->avenrun[2]; |
96 | P_DEBUG("avenrun(float) = %d.%02d / %d.%02d / %d.%02d\n", | 96 | P_DEBUG("avenrun(float) = %d.%02d / %d.%02d / %d.%02d\n", |
97 | LOAD_INT(a0), LOAD_FRAC(a0), LOAD_INT(a1), LOAD_FRAC(a1), | 97 | LOAD_INT(a0), LOAD_FRAC(a0), LOAD_INT(a1), LOAD_FRAC(a1), |
98 | LOAD_INT(a2), LOAD_FRAC(a2)); | 98 | LOAD_INT(a2), LOAD_FRAC(a2)); |
99 | 99 | ||
100 | P_DEBUG("nr_cpus = %u\n", os_data->nr_cpus); | 100 | P_DEBUG("nr_cpus = %u\n", os_data->nr_cpus); |
101 | for (i = 0; i < os_data->nr_cpus; i++) { | 101 | for (i = 0; i < os_data->nr_cpus; i++) { |
102 | P_DEBUG("cpu%u : user = %u, nice = %u, system = %u, " | 102 | P_DEBUG("cpu%u : user = %u, nice = %u, system = %u, " |
103 | "idle = %u, irq = %u, softirq = %u, iowait = %u\n", | 103 | "idle = %u, irq = %u, softirq = %u, iowait = %u\n", |
104 | i, | 104 | i, |
105 | os_data->os_cpu[i].per_cpu_user, | 105 | os_data->os_cpu[i].per_cpu_user, |
106 | os_data->os_cpu[i].per_cpu_nice, | 106 | os_data->os_cpu[i].per_cpu_nice, |
107 | os_data->os_cpu[i].per_cpu_system, | 107 | os_data->os_cpu[i].per_cpu_system, |
108 | os_data->os_cpu[i].per_cpu_idle, | 108 | os_data->os_cpu[i].per_cpu_idle, |
109 | os_data->os_cpu[i].per_cpu_irq, | 109 | os_data->os_cpu[i].per_cpu_irq, |
110 | os_data->os_cpu[i].per_cpu_softirq, | 110 | os_data->os_cpu[i].per_cpu_softirq, |
111 | os_data->os_cpu[i].per_cpu_iowait); | 111 | os_data->os_cpu[i].per_cpu_iowait); |
112 | } | 112 | } |
113 | 113 | ||
114 | P_DEBUG("sync_count_1 = %u\n", os_data->sync_count_1); | 114 | P_DEBUG("sync_count_1 = %u\n", os_data->sync_count_1); |
115 | P_DEBUG("sync_count_2 = %u\n", os_data->sync_count_2); | 115 | P_DEBUG("sync_count_2 = %u\n", os_data->sync_count_2); |
116 | P_DEBUG("timestamp = %lX\n", os_data->timestamp); | 116 | P_DEBUG("timestamp = %lX\n", os_data->timestamp); |
117 | } | 117 | } |
118 | 118 | ||
119 | /* | 119 | /* |
120 | * appldata_get_os_data() | 120 | * appldata_get_os_data() |
121 | * | 121 | * |
122 | * gather OS data | 122 | * gather OS data |
123 | */ | 123 | */ |
124 | static void appldata_get_os_data(void *data) | 124 | static void appldata_get_os_data(void *data) |
125 | { | 125 | { |
126 | int i, j; | 126 | int i, j; |
127 | struct appldata_os_data *os_data; | 127 | struct appldata_os_data *os_data; |
128 | 128 | ||
129 | os_data = data; | 129 | os_data = data; |
130 | os_data->sync_count_1++; | 130 | os_data->sync_count_1++; |
131 | 131 | ||
132 | os_data->nr_cpus = num_online_cpus(); | 132 | os_data->nr_cpus = num_online_cpus(); |
133 | 133 | ||
134 | os_data->nr_threads = nr_threads; | 134 | os_data->nr_threads = nr_threads; |
135 | os_data->nr_running = nr_running(); | 135 | os_data->nr_running = nr_running(); |
136 | os_data->nr_iowait = nr_iowait(); | 136 | os_data->nr_iowait = nr_iowait(); |
137 | os_data->avenrun[0] = avenrun[0] + (FIXED_1/200); | 137 | os_data->avenrun[0] = avenrun[0] + (FIXED_1/200); |
138 | os_data->avenrun[1] = avenrun[1] + (FIXED_1/200); | 138 | os_data->avenrun[1] = avenrun[1] + (FIXED_1/200); |
139 | os_data->avenrun[2] = avenrun[2] + (FIXED_1/200); | 139 | os_data->avenrun[2] = avenrun[2] + (FIXED_1/200); |
140 | 140 | ||
141 | j = 0; | 141 | j = 0; |
142 | for_each_online_cpu(i) { | 142 | for_each_online_cpu(i) { |
143 | os_data->os_cpu[j].per_cpu_user = | 143 | os_data->os_cpu[j].per_cpu_user = |
144 | kstat_cpu(i).cpustat.user; | 144 | cputime_to_jiffies(kstat_cpu(i).cpustat.user); |
145 | os_data->os_cpu[j].per_cpu_nice = | 145 | os_data->os_cpu[j].per_cpu_nice = |
146 | kstat_cpu(i).cpustat.nice; | 146 | cputime_to_jiffies(kstat_cpu(i).cpustat.nice); |
147 | os_data->os_cpu[j].per_cpu_system = | 147 | os_data->os_cpu[j].per_cpu_system = |
148 | kstat_cpu(i).cpustat.system; | 148 | cputime_to_jiffies(kstat_cpu(i).cpustat.system); |
149 | os_data->os_cpu[j].per_cpu_idle = | 149 | os_data->os_cpu[j].per_cpu_idle = |
150 | kstat_cpu(i).cpustat.idle; | 150 | cputime_to_jiffies(kstat_cpu(i).cpustat.idle); |
151 | os_data->os_cpu[j].per_cpu_irq = | 151 | os_data->os_cpu[j].per_cpu_irq = |
152 | kstat_cpu(i).cpustat.irq; | 152 | cputime_to_jiffies(kstat_cpu(i).cpustat.irq); |
153 | os_data->os_cpu[j].per_cpu_softirq = | 153 | os_data->os_cpu[j].per_cpu_softirq = |
154 | kstat_cpu(i).cpustat.softirq; | 154 | cputime_to_jiffies(kstat_cpu(i).cpustat.softirq); |
155 | os_data->os_cpu[j].per_cpu_iowait = | 155 | os_data->os_cpu[j].per_cpu_iowait = |
156 | kstat_cpu(i).cpustat.iowait; | 156 | cputime_to_jiffies(kstat_cpu(i).cpustat.iowait); |
157 | j++; | 157 | j++; |
158 | } | 158 | } |
159 | 159 | ||
160 | os_data->timestamp = get_clock(); | 160 | os_data->timestamp = get_clock(); |
161 | os_data->sync_count_2++; | 161 | os_data->sync_count_2++; |
162 | #ifdef APPLDATA_DEBUG | 162 | #ifdef APPLDATA_DEBUG |
163 | appldata_print_debug(os_data); | 163 | appldata_print_debug(os_data); |
164 | #endif | 164 | #endif |
165 | } | 165 | } |
166 | 166 | ||
167 | 167 | ||
168 | static struct appldata_ops ops = { | 168 | static struct appldata_ops ops = { |
169 | .ctl_nr = CTL_APPLDATA_OS, | 169 | .ctl_nr = CTL_APPLDATA_OS, |
170 | .name = "os", | 170 | .name = "os", |
171 | .record_nr = APPLDATA_RECORD_OS_ID, | 171 | .record_nr = APPLDATA_RECORD_OS_ID, |
172 | .callback = &appldata_get_os_data, | 172 | .callback = &appldata_get_os_data, |
173 | .owner = THIS_MODULE, | 173 | .owner = THIS_MODULE, |
174 | }; | 174 | }; |
175 | 175 | ||
176 | 176 | ||
177 | /* | 177 | /* |
178 | * appldata_os_init() | 178 | * appldata_os_init() |
179 | * | 179 | * |
180 | * init data, register ops | 180 | * init data, register ops |
181 | */ | 181 | */ |
182 | static int __init appldata_os_init(void) | 182 | static int __init appldata_os_init(void) |
183 | { | 183 | { |
184 | int rc, size; | 184 | int rc, size; |
185 | 185 | ||
186 | size = sizeof(struct appldata_os_data) + | 186 | size = sizeof(struct appldata_os_data) + |
187 | (NR_CPUS * sizeof(struct appldata_os_per_cpu)); | 187 | (NR_CPUS * sizeof(struct appldata_os_per_cpu)); |
188 | if (size > APPLDATA_MAX_REC_SIZE) { | 188 | if (size > APPLDATA_MAX_REC_SIZE) { |
189 | P_ERROR("Size of record = %i, bigger than maximum (%i)!\n", | 189 | P_ERROR("Size of record = %i, bigger than maximum (%i)!\n", |
190 | size, APPLDATA_MAX_REC_SIZE); | 190 | size, APPLDATA_MAX_REC_SIZE); |
191 | rc = -ENOMEM; | 191 | rc = -ENOMEM; |
192 | goto out; | 192 | goto out; |
193 | } | 193 | } |
194 | P_DEBUG("sizeof(os) = %i, sizeof(os_cpu) = %lu\n", size, | 194 | P_DEBUG("sizeof(os) = %i, sizeof(os_cpu) = %lu\n", size, |
195 | sizeof(struct appldata_os_per_cpu)); | 195 | sizeof(struct appldata_os_per_cpu)); |
196 | 196 | ||
197 | appldata_os_data = kmalloc(size, GFP_DMA); | 197 | appldata_os_data = kmalloc(size, GFP_DMA); |
198 | if (appldata_os_data == NULL) { | 198 | if (appldata_os_data == NULL) { |
199 | P_ERROR("No memory for %s!\n", ops.name); | 199 | P_ERROR("No memory for %s!\n", ops.name); |
200 | rc = -ENOMEM; | 200 | rc = -ENOMEM; |
201 | goto out; | 201 | goto out; |
202 | } | 202 | } |
203 | memset(appldata_os_data, 0, size); | 203 | memset(appldata_os_data, 0, size); |
204 | 204 | ||
205 | appldata_os_data->per_cpu_size = sizeof(struct appldata_os_per_cpu); | 205 | appldata_os_data->per_cpu_size = sizeof(struct appldata_os_per_cpu); |
206 | appldata_os_data->cpu_offset = offsetof(struct appldata_os_data, | 206 | appldata_os_data->cpu_offset = offsetof(struct appldata_os_data, |
207 | os_cpu); | 207 | os_cpu); |
208 | P_DEBUG("cpu offset = %u\n", appldata_os_data->cpu_offset); | 208 | P_DEBUG("cpu offset = %u\n", appldata_os_data->cpu_offset); |
209 | 209 | ||
210 | ops.data = appldata_os_data; | 210 | ops.data = appldata_os_data; |
211 | ops.size = size; | 211 | ops.size = size; |
212 | rc = appldata_register_ops(&ops); | 212 | rc = appldata_register_ops(&ops); |
213 | if (rc != 0) { | 213 | if (rc != 0) { |
214 | P_ERROR("Error registering ops, rc = %i\n", rc); | 214 | P_ERROR("Error registering ops, rc = %i\n", rc); |
215 | kfree(appldata_os_data); | 215 | kfree(appldata_os_data); |
216 | } else { | 216 | } else { |
217 | P_DEBUG("%s-ops registered!\n", ops.name); | 217 | P_DEBUG("%s-ops registered!\n", ops.name); |
218 | } | 218 | } |
219 | out: | 219 | out: |
220 | return rc; | 220 | return rc; |
221 | } | 221 | } |
222 | 222 | ||
223 | /* | 223 | /* |
224 | * appldata_os_exit() | 224 | * appldata_os_exit() |
225 | * | 225 | * |
226 | * unregister ops | 226 | * unregister ops |
227 | */ | 227 | */ |
228 | static void __exit appldata_os_exit(void) | 228 | static void __exit appldata_os_exit(void) |
229 | { | 229 | { |
230 | appldata_unregister_ops(&ops); | 230 | appldata_unregister_ops(&ops); |
231 | kfree(appldata_os_data); | 231 | kfree(appldata_os_data); |
232 | P_DEBUG("%s-ops unregistered!\n", ops.name); | 232 | P_DEBUG("%s-ops unregistered!\n", ops.name); |
233 | } | 233 | } |
234 | 234 | ||
235 | 235 | ||
236 | module_init(appldata_os_init); | 236 | module_init(appldata_os_init); |
237 | module_exit(appldata_os_exit); | 237 | module_exit(appldata_os_exit); |
238 | 238 | ||
239 | MODULE_LICENSE("GPL"); | 239 | MODULE_LICENSE("GPL"); |
240 | MODULE_AUTHOR("Gerald Schaefer"); | 240 | MODULE_AUTHOR("Gerald Schaefer"); |
241 | MODULE_DESCRIPTION("Linux-VM Monitor Stream, OS statistics"); | 241 | MODULE_DESCRIPTION("Linux-VM Monitor Stream, OS statistics"); |
242 | 242 |
kernel/acct.c
1 | /* | 1 | /* |
2 | * linux/kernel/acct.c | 2 | * linux/kernel/acct.c |
3 | * | 3 | * |
4 | * BSD Process Accounting for Linux | 4 | * BSD Process Accounting for Linux |
5 | * | 5 | * |
6 | * Author: Marco van Wieringen <mvw@planets.elm.net> | 6 | * Author: Marco van Wieringen <mvw@planets.elm.net> |
7 | * | 7 | * |
8 | * Some code based on ideas and code from: | 8 | * Some code based on ideas and code from: |
9 | * Thomas K. Dyas <tdyas@eden.rutgers.edu> | 9 | * Thomas K. Dyas <tdyas@eden.rutgers.edu> |
10 | * | 10 | * |
11 | * This file implements BSD-style process accounting. Whenever any | 11 | * This file implements BSD-style process accounting. Whenever any |
12 | * process exits, an accounting record of type "struct acct" is | 12 | * process exits, an accounting record of type "struct acct" is |
13 | * written to the file specified with the acct() system call. It is | 13 | * written to the file specified with the acct() system call. It is |
14 | * up to user-level programs to do useful things with the accounting | 14 | * up to user-level programs to do useful things with the accounting |
15 | * log. The kernel just provides the raw accounting information. | 15 | * log. The kernel just provides the raw accounting information. |
16 | * | 16 | * |
17 | * (C) Copyright 1995 - 1997 Marco van Wieringen - ELM Consultancy B.V. | 17 | * (C) Copyright 1995 - 1997 Marco van Wieringen - ELM Consultancy B.V. |
18 | * | 18 | * |
19 | * Plugged two leaks. 1) It didn't return acct_file into the free_filps if | 19 | * Plugged two leaks. 1) It didn't return acct_file into the free_filps if |
20 | * the file happened to be read-only. 2) If the accounting was suspended | 20 | * the file happened to be read-only. 2) If the accounting was suspended |
21 | * due to the lack of space it happily allowed to reopen it and completely | 21 | * due to the lack of space it happily allowed to reopen it and completely |
22 | * lost the old acct_file. 3/10/98, Al Viro. | 22 | * lost the old acct_file. 3/10/98, Al Viro. |
23 | * | 23 | * |
24 | * Now we silently close acct_file on attempt to reopen. Cleaned sys_acct(). | 24 | * Now we silently close acct_file on attempt to reopen. Cleaned sys_acct(). |
25 | * XTerms and EMACS are manifestations of pure evil. 21/10/98, AV. | 25 | * XTerms and EMACS are manifestations of pure evil. 21/10/98, AV. |
26 | * | 26 | * |
27 | * Fixed a nasty interaction with with sys_umount(). If the accointing | 27 | * Fixed a nasty interaction with with sys_umount(). If the accointing |
28 | * was suspeneded we failed to stop it on umount(). Messy. | 28 | * was suspeneded we failed to stop it on umount(). Messy. |
29 | * Another one: remount to readonly didn't stop accounting. | 29 | * Another one: remount to readonly didn't stop accounting. |
30 | * Question: what should we do if we have CAP_SYS_ADMIN but not | 30 | * Question: what should we do if we have CAP_SYS_ADMIN but not |
31 | * CAP_SYS_PACCT? Current code does the following: umount returns -EBUSY | 31 | * CAP_SYS_PACCT? Current code does the following: umount returns -EBUSY |
32 | * unless we are messing with the root. In that case we are getting a | 32 | * unless we are messing with the root. In that case we are getting a |
33 | * real mess with do_remount_sb(). 9/11/98, AV. | 33 | * real mess with do_remount_sb(). 9/11/98, AV. |
34 | * | 34 | * |
35 | * Fixed a bunch of races (and pair of leaks). Probably not the best way, | 35 | * Fixed a bunch of races (and pair of leaks). Probably not the best way, |
36 | * but this one obviously doesn't introduce deadlocks. Later. BTW, found | 36 | * but this one obviously doesn't introduce deadlocks. Later. BTW, found |
37 | * one race (and leak) in BSD implementation. | 37 | * one race (and leak) in BSD implementation. |
38 | * OK, that's better. ANOTHER race and leak in BSD variant. There always | 38 | * OK, that's better. ANOTHER race and leak in BSD variant. There always |
39 | * is one more bug... 10/11/98, AV. | 39 | * is one more bug... 10/11/98, AV. |
40 | * | 40 | * |
41 | * Oh, fsck... Oopsable SMP race in do_process_acct() - we must hold | 41 | * Oh, fsck... Oopsable SMP race in do_process_acct() - we must hold |
42 | * ->mmap_sem to walk the vma list of current->mm. Nasty, since it leaks | 42 | * ->mmap_sem to walk the vma list of current->mm. Nasty, since it leaks |
43 | * a struct file opened for write. Fixed. 2/6/2000, AV. | 43 | * a struct file opened for write. Fixed. 2/6/2000, AV. |
44 | */ | 44 | */ |
45 | 45 | ||
46 | #include <linux/config.h> | 46 | #include <linux/config.h> |
47 | #include <linux/mm.h> | 47 | #include <linux/mm.h> |
48 | #include <linux/slab.h> | 48 | #include <linux/slab.h> |
49 | #include <linux/acct.h> | 49 | #include <linux/acct.h> |
50 | #include <linux/file.h> | 50 | #include <linux/file.h> |
51 | #include <linux/tty.h> | 51 | #include <linux/tty.h> |
52 | #include <linux/security.h> | 52 | #include <linux/security.h> |
53 | #include <linux/vfs.h> | 53 | #include <linux/vfs.h> |
54 | #include <linux/jiffies.h> | 54 | #include <linux/jiffies.h> |
55 | #include <linux/times.h> | 55 | #include <linux/times.h> |
56 | #include <linux/syscalls.h> | 56 | #include <linux/syscalls.h> |
57 | #include <linux/mount.h> | 57 | #include <linux/mount.h> |
58 | #include <asm/uaccess.h> | 58 | #include <asm/uaccess.h> |
59 | #include <asm/div64.h> | 59 | #include <asm/div64.h> |
60 | #include <linux/blkdev.h> /* sector_div */ | 60 | #include <linux/blkdev.h> /* sector_div */ |
61 | 61 | ||
62 | /* | 62 | /* |
63 | * These constants control the amount of freespace that suspend and | 63 | * These constants control the amount of freespace that suspend and |
64 | * resume the process accounting system, and the time delay between | 64 | * resume the process accounting system, and the time delay between |
65 | * each check. | 65 | * each check. |
66 | * Turned into sysctl-controllable parameters. AV, 12/11/98 | 66 | * Turned into sysctl-controllable parameters. AV, 12/11/98 |
67 | */ | 67 | */ |
68 | 68 | ||
69 | int acct_parm[3] = {4, 2, 30}; | 69 | int acct_parm[3] = {4, 2, 30}; |
70 | #define RESUME (acct_parm[0]) /* >foo% free space - resume */ | 70 | #define RESUME (acct_parm[0]) /* >foo% free space - resume */ |
71 | #define SUSPEND (acct_parm[1]) /* <foo% free space - suspend */ | 71 | #define SUSPEND (acct_parm[1]) /* <foo% free space - suspend */ |
72 | #define ACCT_TIMEOUT (acct_parm[2]) /* foo second timeout between checks */ | 72 | #define ACCT_TIMEOUT (acct_parm[2]) /* foo second timeout between checks */ |
73 | 73 | ||
74 | /* | 74 | /* |
75 | * External references and all of the globals. | 75 | * External references and all of the globals. |
76 | */ | 76 | */ |
77 | static void do_acct_process(long, struct file *); | 77 | static void do_acct_process(long, struct file *); |
78 | 78 | ||
79 | /* | 79 | /* |
80 | * This structure is used so that all the data protected by lock | 80 | * This structure is used so that all the data protected by lock |
81 | * can be placed in the same cache line as the lock. This primes | 81 | * can be placed in the same cache line as the lock. This primes |
82 | * the cache line to have the data after getting the lock. | 82 | * the cache line to have the data after getting the lock. |
83 | */ | 83 | */ |
84 | struct acct_glbs { | 84 | struct acct_glbs { |
85 | spinlock_t lock; | 85 | spinlock_t lock; |
86 | volatile int active; | 86 | volatile int active; |
87 | volatile int needcheck; | 87 | volatile int needcheck; |
88 | struct file *file; | 88 | struct file *file; |
89 | struct timer_list timer; | 89 | struct timer_list timer; |
90 | }; | 90 | }; |
91 | 91 | ||
92 | static struct acct_glbs acct_globals __cacheline_aligned = {SPIN_LOCK_UNLOCKED}; | 92 | static struct acct_glbs acct_globals __cacheline_aligned = {SPIN_LOCK_UNLOCKED}; |
93 | 93 | ||
94 | /* | 94 | /* |
95 | * Called whenever the timer says to check the free space. | 95 | * Called whenever the timer says to check the free space. |
96 | */ | 96 | */ |
97 | static void acct_timeout(unsigned long unused) | 97 | static void acct_timeout(unsigned long unused) |
98 | { | 98 | { |
99 | acct_globals.needcheck = 1; | 99 | acct_globals.needcheck = 1; |
100 | } | 100 | } |
101 | 101 | ||
102 | /* | 102 | /* |
103 | * Check the amount of free space and suspend/resume accordingly. | 103 | * Check the amount of free space and suspend/resume accordingly. |
104 | */ | 104 | */ |
105 | static int check_free_space(struct file *file) | 105 | static int check_free_space(struct file *file) |
106 | { | 106 | { |
107 | struct kstatfs sbuf; | 107 | struct kstatfs sbuf; |
108 | int res; | 108 | int res; |
109 | int act; | 109 | int act; |
110 | sector_t resume; | 110 | sector_t resume; |
111 | sector_t suspend; | 111 | sector_t suspend; |
112 | 112 | ||
113 | spin_lock(&acct_globals.lock); | 113 | spin_lock(&acct_globals.lock); |
114 | res = acct_globals.active; | 114 | res = acct_globals.active; |
115 | if (!file || !acct_globals.needcheck) | 115 | if (!file || !acct_globals.needcheck) |
116 | goto out; | 116 | goto out; |
117 | spin_unlock(&acct_globals.lock); | 117 | spin_unlock(&acct_globals.lock); |
118 | 118 | ||
119 | /* May block */ | 119 | /* May block */ |
120 | if (vfs_statfs(file->f_dentry->d_inode->i_sb, &sbuf)) | 120 | if (vfs_statfs(file->f_dentry->d_inode->i_sb, &sbuf)) |
121 | return res; | 121 | return res; |
122 | suspend = sbuf.f_blocks * SUSPEND; | 122 | suspend = sbuf.f_blocks * SUSPEND; |
123 | resume = sbuf.f_blocks * RESUME; | 123 | resume = sbuf.f_blocks * RESUME; |
124 | 124 | ||
125 | sector_div(suspend, 100); | 125 | sector_div(suspend, 100); |
126 | sector_div(resume, 100); | 126 | sector_div(resume, 100); |
127 | 127 | ||
128 | if (sbuf.f_bavail <= suspend) | 128 | if (sbuf.f_bavail <= suspend) |
129 | act = -1; | 129 | act = -1; |
130 | else if (sbuf.f_bavail >= resume) | 130 | else if (sbuf.f_bavail >= resume) |
131 | act = 1; | 131 | act = 1; |
132 | else | 132 | else |
133 | act = 0; | 133 | act = 0; |
134 | 134 | ||
135 | /* | 135 | /* |
136 | * If some joker switched acct_globals.file under us we'ld better be | 136 | * If some joker switched acct_globals.file under us we'ld better be |
137 | * silent and _not_ touch anything. | 137 | * silent and _not_ touch anything. |
138 | */ | 138 | */ |
139 | spin_lock(&acct_globals.lock); | 139 | spin_lock(&acct_globals.lock); |
140 | if (file != acct_globals.file) { | 140 | if (file != acct_globals.file) { |
141 | if (act) | 141 | if (act) |
142 | res = act>0; | 142 | res = act>0; |
143 | goto out; | 143 | goto out; |
144 | } | 144 | } |
145 | 145 | ||
146 | if (acct_globals.active) { | 146 | if (acct_globals.active) { |
147 | if (act < 0) { | 147 | if (act < 0) { |
148 | acct_globals.active = 0; | 148 | acct_globals.active = 0; |
149 | printk(KERN_INFO "Process accounting paused\n"); | 149 | printk(KERN_INFO "Process accounting paused\n"); |
150 | } | 150 | } |
151 | } else { | 151 | } else { |
152 | if (act > 0) { | 152 | if (act > 0) { |
153 | acct_globals.active = 1; | 153 | acct_globals.active = 1; |
154 | printk(KERN_INFO "Process accounting resumed\n"); | 154 | printk(KERN_INFO "Process accounting resumed\n"); |
155 | } | 155 | } |
156 | } | 156 | } |
157 | 157 | ||
158 | del_timer(&acct_globals.timer); | 158 | del_timer(&acct_globals.timer); |
159 | acct_globals.needcheck = 0; | 159 | acct_globals.needcheck = 0; |
160 | acct_globals.timer.expires = jiffies + ACCT_TIMEOUT*HZ; | 160 | acct_globals.timer.expires = jiffies + ACCT_TIMEOUT*HZ; |
161 | add_timer(&acct_globals.timer); | 161 | add_timer(&acct_globals.timer); |
162 | res = acct_globals.active; | 162 | res = acct_globals.active; |
163 | out: | 163 | out: |
164 | spin_unlock(&acct_globals.lock); | 164 | spin_unlock(&acct_globals.lock); |
165 | return res; | 165 | return res; |
166 | } | 166 | } |
167 | 167 | ||
168 | /* | 168 | /* |
169 | * Close the old accounting file (if currently open) and then replace | 169 | * Close the old accounting file (if currently open) and then replace |
170 | * it with file (if non-NULL). | 170 | * it with file (if non-NULL). |
171 | * | 171 | * |
172 | * NOTE: acct_globals.lock MUST be held on entry and exit. | 172 | * NOTE: acct_globals.lock MUST be held on entry and exit. |
173 | */ | 173 | */ |
174 | static void acct_file_reopen(struct file *file) | 174 | static void acct_file_reopen(struct file *file) |
175 | { | 175 | { |
176 | struct file *old_acct = NULL; | 176 | struct file *old_acct = NULL; |
177 | 177 | ||
178 | if (acct_globals.file) { | 178 | if (acct_globals.file) { |
179 | old_acct = acct_globals.file; | 179 | old_acct = acct_globals.file; |
180 | del_timer(&acct_globals.timer); | 180 | del_timer(&acct_globals.timer); |
181 | acct_globals.active = 0; | 181 | acct_globals.active = 0; |
182 | acct_globals.needcheck = 0; | 182 | acct_globals.needcheck = 0; |
183 | acct_globals.file = NULL; | 183 | acct_globals.file = NULL; |
184 | } | 184 | } |
185 | if (file) { | 185 | if (file) { |
186 | acct_globals.file = file; | 186 | acct_globals.file = file; |
187 | acct_globals.needcheck = 0; | 187 | acct_globals.needcheck = 0; |
188 | acct_globals.active = 1; | 188 | acct_globals.active = 1; |
189 | /* It's been deleted if it was used before so this is safe */ | 189 | /* It's been deleted if it was used before so this is safe */ |
190 | init_timer(&acct_globals.timer); | 190 | init_timer(&acct_globals.timer); |
191 | acct_globals.timer.function = acct_timeout; | 191 | acct_globals.timer.function = acct_timeout; |
192 | acct_globals.timer.expires = jiffies + ACCT_TIMEOUT*HZ; | 192 | acct_globals.timer.expires = jiffies + ACCT_TIMEOUT*HZ; |
193 | add_timer(&acct_globals.timer); | 193 | add_timer(&acct_globals.timer); |
194 | } | 194 | } |
195 | if (old_acct) { | 195 | if (old_acct) { |
196 | mnt_unpin(old_acct->f_vfsmnt); | 196 | mnt_unpin(old_acct->f_vfsmnt); |
197 | spin_unlock(&acct_globals.lock); | 197 | spin_unlock(&acct_globals.lock); |
198 | do_acct_process(0, old_acct); | 198 | do_acct_process(0, old_acct); |
199 | filp_close(old_acct, NULL); | 199 | filp_close(old_acct, NULL); |
200 | spin_lock(&acct_globals.lock); | 200 | spin_lock(&acct_globals.lock); |
201 | } | 201 | } |
202 | } | 202 | } |
203 | 203 | ||
204 | static int acct_on(char *name) | 204 | static int acct_on(char *name) |
205 | { | 205 | { |
206 | struct file *file; | 206 | struct file *file; |
207 | int error; | 207 | int error; |
208 | 208 | ||
209 | /* Difference from BSD - they don't do O_APPEND */ | 209 | /* Difference from BSD - they don't do O_APPEND */ |
210 | file = filp_open(name, O_WRONLY|O_APPEND|O_LARGEFILE, 0); | 210 | file = filp_open(name, O_WRONLY|O_APPEND|O_LARGEFILE, 0); |
211 | if (IS_ERR(file)) | 211 | if (IS_ERR(file)) |
212 | return PTR_ERR(file); | 212 | return PTR_ERR(file); |
213 | 213 | ||
214 | if (!S_ISREG(file->f_dentry->d_inode->i_mode)) { | 214 | if (!S_ISREG(file->f_dentry->d_inode->i_mode)) { |
215 | filp_close(file, NULL); | 215 | filp_close(file, NULL); |
216 | return -EACCES; | 216 | return -EACCES; |
217 | } | 217 | } |
218 | 218 | ||
219 | if (!file->f_op->write) { | 219 | if (!file->f_op->write) { |
220 | filp_close(file, NULL); | 220 | filp_close(file, NULL); |
221 | return -EIO; | 221 | return -EIO; |
222 | } | 222 | } |
223 | 223 | ||
224 | error = security_acct(file); | 224 | error = security_acct(file); |
225 | if (error) { | 225 | if (error) { |
226 | filp_close(file, NULL); | 226 | filp_close(file, NULL); |
227 | return error; | 227 | return error; |
228 | } | 228 | } |
229 | 229 | ||
230 | spin_lock(&acct_globals.lock); | 230 | spin_lock(&acct_globals.lock); |
231 | mnt_pin(file->f_vfsmnt); | 231 | mnt_pin(file->f_vfsmnt); |
232 | acct_file_reopen(file); | 232 | acct_file_reopen(file); |
233 | spin_unlock(&acct_globals.lock); | 233 | spin_unlock(&acct_globals.lock); |
234 | 234 | ||
235 | mntput(file->f_vfsmnt); /* it's pinned, now give up active reference */ | 235 | mntput(file->f_vfsmnt); /* it's pinned, now give up active reference */ |
236 | 236 | ||
237 | return 0; | 237 | return 0; |
238 | } | 238 | } |
239 | 239 | ||
240 | /** | 240 | /** |
241 | * sys_acct - enable/disable process accounting | 241 | * sys_acct - enable/disable process accounting |
242 | * @name: file name for accounting records or NULL to shutdown accounting | 242 | * @name: file name for accounting records or NULL to shutdown accounting |
243 | * | 243 | * |
244 | * Returns 0 for success or negative errno values for failure. | 244 | * Returns 0 for success or negative errno values for failure. |
245 | * | 245 | * |
246 | * sys_acct() is the only system call needed to implement process | 246 | * sys_acct() is the only system call needed to implement process |
247 | * accounting. It takes the name of the file where accounting records | 247 | * accounting. It takes the name of the file where accounting records |
248 | * should be written. If the filename is NULL, accounting will be | 248 | * should be written. If the filename is NULL, accounting will be |
249 | * shutdown. | 249 | * shutdown. |
250 | */ | 250 | */ |
251 | asmlinkage long sys_acct(const char __user *name) | 251 | asmlinkage long sys_acct(const char __user *name) |
252 | { | 252 | { |
253 | int error; | 253 | int error; |
254 | 254 | ||
255 | if (!capable(CAP_SYS_PACCT)) | 255 | if (!capable(CAP_SYS_PACCT)) |
256 | return -EPERM; | 256 | return -EPERM; |
257 | 257 | ||
258 | if (name) { | 258 | if (name) { |
259 | char *tmp = getname(name); | 259 | char *tmp = getname(name); |
260 | if (IS_ERR(tmp)) | 260 | if (IS_ERR(tmp)) |
261 | return (PTR_ERR(tmp)); | 261 | return (PTR_ERR(tmp)); |
262 | error = acct_on(tmp); | 262 | error = acct_on(tmp); |
263 | putname(tmp); | 263 | putname(tmp); |
264 | } else { | 264 | } else { |
265 | error = security_acct(NULL); | 265 | error = security_acct(NULL); |
266 | if (!error) { | 266 | if (!error) { |
267 | spin_lock(&acct_globals.lock); | 267 | spin_lock(&acct_globals.lock); |
268 | acct_file_reopen(NULL); | 268 | acct_file_reopen(NULL); |
269 | spin_unlock(&acct_globals.lock); | 269 | spin_unlock(&acct_globals.lock); |
270 | } | 270 | } |
271 | } | 271 | } |
272 | return error; | 272 | return error; |
273 | } | 273 | } |
274 | 274 | ||
275 | /** | 275 | /** |
276 | * acct_auto_close - turn off a filesystem's accounting if it is on | 276 | * acct_auto_close - turn off a filesystem's accounting if it is on |
277 | * @m: vfsmount being shut down | 277 | * @m: vfsmount being shut down |
278 | * | 278 | * |
279 | * If the accounting is turned on for a file in the subtree pointed to | 279 | * If the accounting is turned on for a file in the subtree pointed to |
280 | * to by m, turn accounting off. Done when m is about to die. | 280 | * to by m, turn accounting off. Done when m is about to die. |
281 | */ | 281 | */ |
282 | void acct_auto_close_mnt(struct vfsmount *m) | 282 | void acct_auto_close_mnt(struct vfsmount *m) |
283 | { | 283 | { |
284 | spin_lock(&acct_globals.lock); | 284 | spin_lock(&acct_globals.lock); |
285 | if (acct_globals.file && acct_globals.file->f_vfsmnt == m) | 285 | if (acct_globals.file && acct_globals.file->f_vfsmnt == m) |
286 | acct_file_reopen(NULL); | 286 | acct_file_reopen(NULL); |
287 | spin_unlock(&acct_globals.lock); | 287 | spin_unlock(&acct_globals.lock); |
288 | } | 288 | } |
289 | 289 | ||
290 | /** | 290 | /** |
291 | * acct_auto_close - turn off a filesystem's accounting if it is on | 291 | * acct_auto_close - turn off a filesystem's accounting if it is on |
292 | * @sb: super block for the filesystem | 292 | * @sb: super block for the filesystem |
293 | * | 293 | * |
294 | * If the accounting is turned on for a file in the filesystem pointed | 294 | * If the accounting is turned on for a file in the filesystem pointed |
295 | * to by sb, turn accounting off. | 295 | * to by sb, turn accounting off. |
296 | */ | 296 | */ |
297 | void acct_auto_close(struct super_block *sb) | 297 | void acct_auto_close(struct super_block *sb) |
298 | { | 298 | { |
299 | spin_lock(&acct_globals.lock); | 299 | spin_lock(&acct_globals.lock); |
300 | if (acct_globals.file && | 300 | if (acct_globals.file && |
301 | acct_globals.file->f_vfsmnt->mnt_sb == sb) { | 301 | acct_globals.file->f_vfsmnt->mnt_sb == sb) { |
302 | acct_file_reopen(NULL); | 302 | acct_file_reopen(NULL); |
303 | } | 303 | } |
304 | spin_unlock(&acct_globals.lock); | 304 | spin_unlock(&acct_globals.lock); |
305 | } | 305 | } |
306 | 306 | ||
307 | /* | 307 | /* |
308 | * encode an unsigned long into a comp_t | 308 | * encode an unsigned long into a comp_t |
309 | * | 309 | * |
310 | * This routine has been adopted from the encode_comp_t() function in | 310 | * This routine has been adopted from the encode_comp_t() function in |
311 | * the kern_acct.c file of the FreeBSD operating system. The encoding | 311 | * the kern_acct.c file of the FreeBSD operating system. The encoding |
312 | * is a 13-bit fraction with a 3-bit (base 8) exponent. | 312 | * is a 13-bit fraction with a 3-bit (base 8) exponent. |
313 | */ | 313 | */ |
314 | 314 | ||
315 | #define MANTSIZE 13 /* 13 bit mantissa. */ | 315 | #define MANTSIZE 13 /* 13 bit mantissa. */ |
316 | #define EXPSIZE 3 /* Base 8 (3 bit) exponent. */ | 316 | #define EXPSIZE 3 /* Base 8 (3 bit) exponent. */ |
317 | #define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */ | 317 | #define MAXFRACT ((1 << MANTSIZE) - 1) /* Maximum fractional value. */ |
318 | 318 | ||
319 | static comp_t encode_comp_t(unsigned long value) | 319 | static comp_t encode_comp_t(unsigned long value) |
320 | { | 320 | { |
321 | int exp, rnd; | 321 | int exp, rnd; |
322 | 322 | ||
323 | exp = rnd = 0; | 323 | exp = rnd = 0; |
324 | while (value > MAXFRACT) { | 324 | while (value > MAXFRACT) { |
325 | rnd = value & (1 << (EXPSIZE - 1)); /* Round up? */ | 325 | rnd = value & (1 << (EXPSIZE - 1)); /* Round up? */ |
326 | value >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */ | 326 | value >>= EXPSIZE; /* Base 8 exponent == 3 bit shift. */ |
327 | exp++; | 327 | exp++; |
328 | } | 328 | } |
329 | 329 | ||
330 | /* | 330 | /* |
331 | * If we need to round up, do it (and handle overflow correctly). | 331 | * If we need to round up, do it (and handle overflow correctly). |
332 | */ | 332 | */ |
333 | if (rnd && (++value > MAXFRACT)) { | 333 | if (rnd && (++value > MAXFRACT)) { |
334 | value >>= EXPSIZE; | 334 | value >>= EXPSIZE; |
335 | exp++; | 335 | exp++; |
336 | } | 336 | } |
337 | 337 | ||
338 | /* | 338 | /* |
339 | * Clean it up and polish it off. | 339 | * Clean it up and polish it off. |
340 | */ | 340 | */ |
341 | exp <<= MANTSIZE; /* Shift the exponent into place */ | 341 | exp <<= MANTSIZE; /* Shift the exponent into place */ |
342 | exp += value; /* and add on the mantissa. */ | 342 | exp += value; /* and add on the mantissa. */ |
343 | return exp; | 343 | return exp; |
344 | } | 344 | } |
345 | 345 | ||
346 | #if ACCT_VERSION==1 || ACCT_VERSION==2 | 346 | #if ACCT_VERSION==1 || ACCT_VERSION==2 |
347 | /* | 347 | /* |
348 | * encode an u64 into a comp2_t (24 bits) | 348 | * encode an u64 into a comp2_t (24 bits) |
349 | * | 349 | * |
350 | * Format: 5 bit base 2 exponent, 20 bits mantissa. | 350 | * Format: 5 bit base 2 exponent, 20 bits mantissa. |
351 | * The leading bit of the mantissa is not stored, but implied for | 351 | * The leading bit of the mantissa is not stored, but implied for |
352 | * non-zero exponents. | 352 | * non-zero exponents. |
353 | * Largest encodable value is 50 bits. | 353 | * Largest encodable value is 50 bits. |
354 | */ | 354 | */ |
355 | 355 | ||
356 | #define MANTSIZE2 20 /* 20 bit mantissa. */ | 356 | #define MANTSIZE2 20 /* 20 bit mantissa. */ |
357 | #define EXPSIZE2 5 /* 5 bit base 2 exponent. */ | 357 | #define EXPSIZE2 5 /* 5 bit base 2 exponent. */ |
358 | #define MAXFRACT2 ((1ul << MANTSIZE2) - 1) /* Maximum fractional value. */ | 358 | #define MAXFRACT2 ((1ul << MANTSIZE2) - 1) /* Maximum fractional value. */ |
359 | #define MAXEXP2 ((1 <<EXPSIZE2) - 1) /* Maximum exponent. */ | 359 | #define MAXEXP2 ((1 <<EXPSIZE2) - 1) /* Maximum exponent. */ |
360 | 360 | ||
361 | static comp2_t encode_comp2_t(u64 value) | 361 | static comp2_t encode_comp2_t(u64 value) |
362 | { | 362 | { |
363 | int exp, rnd; | 363 | int exp, rnd; |
364 | 364 | ||
365 | exp = (value > (MAXFRACT2>>1)); | 365 | exp = (value > (MAXFRACT2>>1)); |
366 | rnd = 0; | 366 | rnd = 0; |
367 | while (value > MAXFRACT2) { | 367 | while (value > MAXFRACT2) { |
368 | rnd = value & 1; | 368 | rnd = value & 1; |
369 | value >>= 1; | 369 | value >>= 1; |
370 | exp++; | 370 | exp++; |
371 | } | 371 | } |
372 | 372 | ||
373 | /* | 373 | /* |
374 | * If we need to round up, do it (and handle overflow correctly). | 374 | * If we need to round up, do it (and handle overflow correctly). |
375 | */ | 375 | */ |
376 | if (rnd && (++value > MAXFRACT2)) { | 376 | if (rnd && (++value > MAXFRACT2)) { |
377 | value >>= 1; | 377 | value >>= 1; |
378 | exp++; | 378 | exp++; |
379 | } | 379 | } |
380 | 380 | ||
381 | if (exp > MAXEXP2) { | 381 | if (exp > MAXEXP2) { |
382 | /* Overflow. Return largest representable number instead. */ | 382 | /* Overflow. Return largest representable number instead. */ |
383 | return (1ul << (MANTSIZE2+EXPSIZE2-1)) - 1; | 383 | return (1ul << (MANTSIZE2+EXPSIZE2-1)) - 1; |
384 | } else { | 384 | } else { |
385 | return (value & (MAXFRACT2>>1)) | (exp << (MANTSIZE2-1)); | 385 | return (value & (MAXFRACT2>>1)) | (exp << (MANTSIZE2-1)); |
386 | } | 386 | } |
387 | } | 387 | } |
388 | #endif | 388 | #endif |
389 | 389 | ||
390 | #if ACCT_VERSION==3 | 390 | #if ACCT_VERSION==3 |
391 | /* | 391 | /* |
392 | * encode an u64 into a 32 bit IEEE float | 392 | * encode an u64 into a 32 bit IEEE float |
393 | */ | 393 | */ |
394 | static u32 encode_float(u64 value) | 394 | static u32 encode_float(u64 value) |
395 | { | 395 | { |
396 | unsigned exp = 190; | 396 | unsigned exp = 190; |
397 | unsigned u; | 397 | unsigned u; |
398 | 398 | ||
399 | if (value==0) return 0; | 399 | if (value==0) return 0; |
400 | while ((s64)value > 0){ | 400 | while ((s64)value > 0){ |
401 | value <<= 1; | 401 | value <<= 1; |
402 | exp--; | 402 | exp--; |
403 | } | 403 | } |
404 | u = (u32)(value >> 40) & 0x7fffffu; | 404 | u = (u32)(value >> 40) & 0x7fffffu; |
405 | return u | (exp << 23); | 405 | return u | (exp << 23); |
406 | } | 406 | } |
407 | #endif | 407 | #endif |
408 | 408 | ||
409 | /* | 409 | /* |
410 | * Write an accounting entry for an exiting process | 410 | * Write an accounting entry for an exiting process |
411 | * | 411 | * |
412 | * The acct_process() call is the workhorse of the process | 412 | * The acct_process() call is the workhorse of the process |
413 | * accounting system. The struct acct is built here and then written | 413 | * accounting system. The struct acct is built here and then written |
414 | * into the accounting file. This function should only be called from | 414 | * into the accounting file. This function should only be called from |
415 | * do_exit(). | 415 | * do_exit(). |
416 | */ | 416 | */ |
417 | 417 | ||
418 | /* | 418 | /* |
419 | * do_acct_process does all actual work. Caller holds the reference to file. | 419 | * do_acct_process does all actual work. Caller holds the reference to file. |
420 | */ | 420 | */ |
421 | static void do_acct_process(long exitcode, struct file *file) | 421 | static void do_acct_process(long exitcode, struct file *file) |
422 | { | 422 | { |
423 | acct_t ac; | 423 | acct_t ac; |
424 | mm_segment_t fs; | 424 | mm_segment_t fs; |
425 | unsigned long vsize; | 425 | unsigned long vsize; |
426 | unsigned long flim; | 426 | unsigned long flim; |
427 | u64 elapsed; | 427 | u64 elapsed; |
428 | u64 run_time; | 428 | u64 run_time; |
429 | struct timespec uptime; | 429 | struct timespec uptime; |
430 | unsigned long jiffies; | ||
430 | 431 | ||
431 | /* | 432 | /* |
432 | * First check to see if there is enough free_space to continue | 433 | * First check to see if there is enough free_space to continue |
433 | * the process accounting system. | 434 | * the process accounting system. |
434 | */ | 435 | */ |
435 | if (!check_free_space(file)) | 436 | if (!check_free_space(file)) |
436 | return; | 437 | return; |
437 | 438 | ||
438 | /* | 439 | /* |
439 | * Fill the accounting struct with the needed info as recorded | 440 | * Fill the accounting struct with the needed info as recorded |
440 | * by the different kernel functions. | 441 | * by the different kernel functions. |
441 | */ | 442 | */ |
442 | memset((caddr_t)&ac, 0, sizeof(acct_t)); | 443 | memset((caddr_t)&ac, 0, sizeof(acct_t)); |
443 | 444 | ||
444 | ac.ac_version = ACCT_VERSION | ACCT_BYTEORDER; | 445 | ac.ac_version = ACCT_VERSION | ACCT_BYTEORDER; |
445 | strlcpy(ac.ac_comm, current->comm, sizeof(ac.ac_comm)); | 446 | strlcpy(ac.ac_comm, current->comm, sizeof(ac.ac_comm)); |
446 | 447 | ||
447 | /* calculate run_time in nsec*/ | 448 | /* calculate run_time in nsec*/ |
448 | do_posix_clock_monotonic_gettime(&uptime); | 449 | do_posix_clock_monotonic_gettime(&uptime); |
449 | run_time = (u64)uptime.tv_sec*NSEC_PER_SEC + uptime.tv_nsec; | 450 | run_time = (u64)uptime.tv_sec*NSEC_PER_SEC + uptime.tv_nsec; |
450 | run_time -= (u64)current->start_time.tv_sec*NSEC_PER_SEC | 451 | run_time -= (u64)current->start_time.tv_sec*NSEC_PER_SEC |
451 | + current->start_time.tv_nsec; | 452 | + current->start_time.tv_nsec; |
452 | /* convert nsec -> AHZ */ | 453 | /* convert nsec -> AHZ */ |
453 | elapsed = nsec_to_AHZ(run_time); | 454 | elapsed = nsec_to_AHZ(run_time); |
454 | #if ACCT_VERSION==3 | 455 | #if ACCT_VERSION==3 |
455 | ac.ac_etime = encode_float(elapsed); | 456 | ac.ac_etime = encode_float(elapsed); |
456 | #else | 457 | #else |
457 | ac.ac_etime = encode_comp_t(elapsed < (unsigned long) -1l ? | 458 | ac.ac_etime = encode_comp_t(elapsed < (unsigned long) -1l ? |
458 | (unsigned long) elapsed : (unsigned long) -1l); | 459 | (unsigned long) elapsed : (unsigned long) -1l); |
459 | #endif | 460 | #endif |
460 | #if ACCT_VERSION==1 || ACCT_VERSION==2 | 461 | #if ACCT_VERSION==1 || ACCT_VERSION==2 |
461 | { | 462 | { |
462 | /* new enlarged etime field */ | 463 | /* new enlarged etime field */ |
463 | comp2_t etime = encode_comp2_t(elapsed); | 464 | comp2_t etime = encode_comp2_t(elapsed); |
464 | ac.ac_etime_hi = etime >> 16; | 465 | ac.ac_etime_hi = etime >> 16; |
465 | ac.ac_etime_lo = (u16) etime; | 466 | ac.ac_etime_lo = (u16) etime; |
466 | } | 467 | } |
467 | #endif | 468 | #endif |
468 | do_div(elapsed, AHZ); | 469 | do_div(elapsed, AHZ); |
469 | ac.ac_btime = xtime.tv_sec - elapsed; | 470 | ac.ac_btime = xtime.tv_sec - elapsed; |
470 | ac.ac_utime = encode_comp_t(jiffies_to_AHZ( | 471 | jiffies = cputime_to_jiffies(cputime_add(current->group_leader->utime, |
471 | current->signal->utime + | 472 | current->signal->utime)); |
472 | current->group_leader->utime)); | 473 | ac.ac_utime = encode_comp_t(jiffies_to_AHZ(jiffies)); |
473 | ac.ac_stime = encode_comp_t(jiffies_to_AHZ( | 474 | jiffies = cputime_to_jiffies(cputime_add(current->group_leader->stime, |
474 | current->signal->stime + | 475 | current->signal->stime)); |
475 | current->group_leader->stime)); | 476 | ac.ac_stime = encode_comp_t(jiffies_to_AHZ(jiffies)); |
476 | /* we really need to bite the bullet and change layout */ | 477 | /* we really need to bite the bullet and change layout */ |
477 | ac.ac_uid = current->uid; | 478 | ac.ac_uid = current->uid; |
478 | ac.ac_gid = current->gid; | 479 | ac.ac_gid = current->gid; |
479 | #if ACCT_VERSION==2 | 480 | #if ACCT_VERSION==2 |
480 | ac.ac_ahz = AHZ; | 481 | ac.ac_ahz = AHZ; |
481 | #endif | 482 | #endif |
482 | #if ACCT_VERSION==1 || ACCT_VERSION==2 | 483 | #if ACCT_VERSION==1 || ACCT_VERSION==2 |
483 | /* backward-compatible 16 bit fields */ | 484 | /* backward-compatible 16 bit fields */ |
484 | ac.ac_uid16 = current->uid; | 485 | ac.ac_uid16 = current->uid; |
485 | ac.ac_gid16 = current->gid; | 486 | ac.ac_gid16 = current->gid; |
486 | #endif | 487 | #endif |
487 | #if ACCT_VERSION==3 | 488 | #if ACCT_VERSION==3 |
488 | ac.ac_pid = current->tgid; | 489 | ac.ac_pid = current->tgid; |
489 | ac.ac_ppid = current->parent->tgid; | 490 | ac.ac_ppid = current->parent->tgid; |
490 | #endif | 491 | #endif |
491 | 492 | ||
492 | read_lock(&tasklist_lock); /* pin current->signal */ | 493 | read_lock(&tasklist_lock); /* pin current->signal */ |
493 | ac.ac_tty = current->signal->tty ? | 494 | ac.ac_tty = current->signal->tty ? |
494 | old_encode_dev(tty_devnum(current->signal->tty)) : 0; | 495 | old_encode_dev(tty_devnum(current->signal->tty)) : 0; |
495 | read_unlock(&tasklist_lock); | 496 | read_unlock(&tasklist_lock); |
496 | 497 | ||
497 | ac.ac_flag = 0; | 498 | ac.ac_flag = 0; |
498 | if (current->flags & PF_FORKNOEXEC) | 499 | if (current->flags & PF_FORKNOEXEC) |
499 | ac.ac_flag |= AFORK; | 500 | ac.ac_flag |= AFORK; |
500 | if (current->flags & PF_SUPERPRIV) | 501 | if (current->flags & PF_SUPERPRIV) |
501 | ac.ac_flag |= ASU; | 502 | ac.ac_flag |= ASU; |
502 | if (current->flags & PF_DUMPCORE) | 503 | if (current->flags & PF_DUMPCORE) |
503 | ac.ac_flag |= ACORE; | 504 | ac.ac_flag |= ACORE; |
504 | if (current->flags & PF_SIGNALED) | 505 | if (current->flags & PF_SIGNALED) |
505 | ac.ac_flag |= AXSIG; | 506 | ac.ac_flag |= AXSIG; |
506 | 507 | ||
507 | vsize = 0; | 508 | vsize = 0; |
508 | if (current->mm) { | 509 | if (current->mm) { |
509 | struct vm_area_struct *vma; | 510 | struct vm_area_struct *vma; |
510 | down_read(¤t->mm->mmap_sem); | 511 | down_read(¤t->mm->mmap_sem); |
511 | vma = current->mm->mmap; | 512 | vma = current->mm->mmap; |
512 | while (vma) { | 513 | while (vma) { |
513 | vsize += vma->vm_end - vma->vm_start; | 514 | vsize += vma->vm_end - vma->vm_start; |
514 | vma = vma->vm_next; | 515 | vma = vma->vm_next; |
515 | } | 516 | } |
516 | up_read(¤t->mm->mmap_sem); | 517 | up_read(¤t->mm->mmap_sem); |
517 | } | 518 | } |
518 | vsize = vsize / 1024; | 519 | vsize = vsize / 1024; |
519 | ac.ac_mem = encode_comp_t(vsize); | 520 | ac.ac_mem = encode_comp_t(vsize); |
520 | ac.ac_io = encode_comp_t(0 /* current->io_usage */); /* %% */ | 521 | ac.ac_io = encode_comp_t(0 /* current->io_usage */); /* %% */ |
521 | ac.ac_rw = encode_comp_t(ac.ac_io / 1024); | 522 | ac.ac_rw = encode_comp_t(ac.ac_io / 1024); |
522 | ac.ac_minflt = encode_comp_t(current->signal->min_flt + | 523 | ac.ac_minflt = encode_comp_t(current->signal->min_flt + |
523 | current->group_leader->min_flt); | 524 | current->group_leader->min_flt); |
524 | ac.ac_majflt = encode_comp_t(current->signal->maj_flt + | 525 | ac.ac_majflt = encode_comp_t(current->signal->maj_flt + |
525 | current->group_leader->maj_flt); | 526 | current->group_leader->maj_flt); |
526 | ac.ac_swaps = encode_comp_t(0); | 527 | ac.ac_swaps = encode_comp_t(0); |
527 | ac.ac_exitcode = exitcode; | 528 | ac.ac_exitcode = exitcode; |
528 | 529 | ||
529 | /* | 530 | /* |
530 | * Kernel segment override to datasegment and write it | 531 | * Kernel segment override to datasegment and write it |
531 | * to the accounting file. | 532 | * to the accounting file. |
532 | */ | 533 | */ |
533 | fs = get_fs(); | 534 | fs = get_fs(); |
534 | set_fs(KERNEL_DS); | 535 | set_fs(KERNEL_DS); |
535 | /* | 536 | /* |
536 | * Accounting records are not subject to resource limits. | 537 | * Accounting records are not subject to resource limits. |
537 | */ | 538 | */ |
538 | flim = current->signal->rlim[RLIMIT_FSIZE].rlim_cur; | 539 | flim = current->signal->rlim[RLIMIT_FSIZE].rlim_cur; |
539 | current->signal->rlim[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; | 540 | current->signal->rlim[RLIMIT_FSIZE].rlim_cur = RLIM_INFINITY; |
540 | file->f_op->write(file, (char *)&ac, | 541 | file->f_op->write(file, (char *)&ac, |
541 | sizeof(acct_t), &file->f_pos); | 542 | sizeof(acct_t), &file->f_pos); |
542 | current->signal->rlim[RLIMIT_FSIZE].rlim_cur = flim; | 543 | current->signal->rlim[RLIMIT_FSIZE].rlim_cur = flim; |
543 | set_fs(fs); | 544 | set_fs(fs); |
544 | } | 545 | } |
545 | 546 | ||
546 | /** | 547 | /** |
547 | * acct_process - now just a wrapper around do_acct_process | 548 | * acct_process - now just a wrapper around do_acct_process |
548 | * @exitcode: task exit code | 549 | * @exitcode: task exit code |
549 | * | 550 | * |
550 | * handles process accounting for an exiting task | 551 | * handles process accounting for an exiting task |
551 | */ | 552 | */ |
552 | void acct_process(long exitcode) | 553 | void acct_process(long exitcode) |
553 | { | 554 | { |
554 | struct file *file = NULL; | 555 | struct file *file = NULL; |
555 | 556 | ||
556 | /* | 557 | /* |
557 | * accelerate the common fastpath: | 558 | * accelerate the common fastpath: |
558 | */ | 559 | */ |
559 | if (!acct_globals.file) | 560 | if (!acct_globals.file) |
560 | return; | 561 | return; |
561 | 562 | ||
562 | spin_lock(&acct_globals.lock); | 563 | spin_lock(&acct_globals.lock); |
563 | file = acct_globals.file; | 564 | file = acct_globals.file; |
564 | if (unlikely(!file)) { | 565 | if (unlikely(!file)) { |
565 | spin_unlock(&acct_globals.lock); | 566 | spin_unlock(&acct_globals.lock); |
566 | return; | 567 | return; |
567 | } | 568 | } |
568 | get_file(file); | 569 | get_file(file); |
569 | spin_unlock(&acct_globals.lock); | 570 | spin_unlock(&acct_globals.lock); |
570 | 571 | ||
571 | do_acct_process(exitcode, file); | 572 | do_acct_process(exitcode, file); |
572 | fput(file); | 573 | fput(file); |
573 | } | 574 | } |
574 | 575 | ||
575 | 576 | ||
576 | /** | 577 | /** |
577 | * acct_update_integrals - update mm integral fields in task_struct | 578 | * acct_update_integrals - update mm integral fields in task_struct |
578 | * @tsk: task_struct for accounting | 579 | * @tsk: task_struct for accounting |
579 | */ | 580 | */ |
580 | void acct_update_integrals(struct task_struct *tsk) | 581 | void acct_update_integrals(struct task_struct *tsk) |
581 | { | 582 | { |
582 | if (likely(tsk->mm)) { | 583 | if (likely(tsk->mm)) { |
583 | long delta = tsk->stime - tsk->acct_stimexpd; | 584 | long delta = |
585 | cputime_to_jiffies(tsk->stime) - tsk->acct_stimexpd; | ||
584 | 586 | ||
585 | if (delta == 0) | 587 | if (delta == 0) |
586 | return; | 588 | return; |
587 | tsk->acct_stimexpd = tsk->stime; | 589 | tsk->acct_stimexpd = tsk->stime; |
588 | tsk->acct_rss_mem1 += delta * get_mm_rss(tsk->mm); | 590 | tsk->acct_rss_mem1 += delta * get_mm_rss(tsk->mm); |
589 | tsk->acct_vm_mem1 += delta * tsk->mm->total_vm; | 591 | tsk->acct_vm_mem1 += delta * tsk->mm->total_vm; |
590 | } | 592 | } |
591 | } | 593 | } |
592 | 594 | ||
593 | /** | 595 | /** |
594 | * acct_clear_integrals - clear the mm integral fields in task_struct | 596 | * acct_clear_integrals - clear the mm integral fields in task_struct |
595 | * @tsk: task_struct whose accounting fields are cleared | 597 | * @tsk: task_struct whose accounting fields are cleared |
596 | */ | 598 | */ |
597 | void acct_clear_integrals(struct task_struct *tsk) | 599 | void acct_clear_integrals(struct task_struct *tsk) |
598 | { | 600 | { |
599 | if (tsk) { | 601 | if (tsk) { |
600 | tsk->acct_stimexpd = 0; | 602 | tsk->acct_stimexpd = 0; |
601 | tsk->acct_rss_mem1 = 0; | 603 | tsk->acct_rss_mem1 = 0; |
602 | tsk->acct_vm_mem1 = 0; | 604 | tsk->acct_vm_mem1 = 0; |
603 | } | 605 | } |
604 | } | 606 | } |
605 | 607 |