1 /* Copyright 2005 Princeton University
3 Redistribution and use in source and binary forms, with or without
4 modification, are permitted provided that the following conditions
7 * Redistributions of source code must retain the above copyright
8 notice, this list of conditions and the following disclaimer.
10 * Redistributions in binary form must reproduce the above
11 copyright notice, this list of conditions and the following
12 disclaimer in the documentation and/or other materials provided
13 with the distribution.
15 * Neither the name of the copyright holder nor the names of its
16 contributors may be used to endorse or promote products derived
17 from this software without specific prior written permission.
19 THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
20 "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
21 LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
22 A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL PRINCETON
23 UNIVERSITY OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
24 INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING,
25 BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS
26 OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
27 AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY
29 WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
30 POSSIBILITY OF SUCH DAMAGE.
45 #include <sys/resource.h>
49 #include "planetlab.h"
52 create_context(xid_t ctx, uint64_t bcaps, struct sliver_resources *slr)
54 struct vc_ctx_caps vc_caps;
55 struct vc_net_nx vc_net;
56 struct vc_net_flags vc_nf;
58 /* Create network context */
59 if (vc_net_create(ctx) == VC_NOCTX)
62 /* Make the network context persistent */
63 vc_nf.mask = vc_nf.flagword = VC_NXF_PERSISTENT;
64 if (vc_set_nflags(ctx, &vc_nf))
67 /* XXX: Allow access to all IPv4 addresses (for now) */
68 vc_net.type = vcNET_IPV4;
72 if (vc_net_add(ctx, &vc_net) == -1)
76 * Create context info - this sets the STATE_SETUP and STATE_INIT flags.
78 if (vc_ctx_create(ctx) == VC_NOCTX)
81 /* Set capabilities - these don't take effect until SETUP flag is unset */
82 vc_caps.bcaps = bcaps;
83 vc_caps.bmask = ~0ULL; /* currently unused */
84 vc_caps.ccaps = 0; /* don't want any of these */
85 vc_caps.cmask = ~0ULL;
86 if (vc_set_ccaps(ctx, &vc_caps))
89 pl_set_limits(ctx, slr);
95 pl_setup_done(xid_t ctx)
97 struct vc_ctx_flags vc_flags;
99 /* unset SETUP flag - this allows other processes to migrate */
100 /* set the PERSISTENT flag - so the context doesn't vanish */
101 /* Don't clear the STATE_INIT flag, as that would make us the init task. */
102 vc_flags.mask = VC_VXF_STATE_SETUP|VC_VXF_PERSISTENT;
103 vc_flags.flagword = VC_VXF_PERSISTENT;
104 if (vc_set_cflags(ctx, &vc_flags))
110 #define RETRY_LIMIT 10
113 pl_chcontext(xid_t ctx, uint64_t bcaps, struct sliver_resources *slr)
116 int net_migrated = 0;
120 struct vc_ctx_flags vc_flags;
122 if (vc_get_cflags(ctx, &vc_flags))
127 /* context doesn't exist - create it */
128 if (create_context(ctx, bcaps,slr))
131 /* another process beat us in a race */
134 /* another process is creating - poll the SETUP flag */
139 /* created context and migrated to it i.e., we're done */
143 /* check the SETUP flag */
144 if (vc_flags.flagword & VC_VXF_STATE_SETUP)
146 /* context is still being setup - wait a while then retry */
147 if (retry_count++ >= RETRY_LIMIT)
156 /* context has been setup */
158 if (net_migrated || !vc_net_migrate(ctx))
160 if (!vc_ctx_migrate(ctx, 0))
165 /* context disappeared - retry */
171 /* it's okay for a syscall to fail because the context doesn't exist */
172 #define VC_SYSCALL(x) \
176 return errno == ESRCH ? 0 : -1; \
181 pl_setsched(xid_t ctx, uint32_t cpu_share, uint32_t cpu_sched_flags)
183 struct vc_set_sched vc_sched;
184 struct vc_ctx_flags vc_flags;
187 vc_sched.set_mask = (VC_VXSM_FILL_RATE | VC_VXSM_INTERVAL | VC_VXSM_TOKENS |
188 VC_VXSM_TOKENS_MIN | VC_VXSM_TOKENS_MAX | VC_VXSM_MSEC |
189 VC_VXSM_FILL_RATE2 | VC_VXSM_INTERVAL2 | VC_VXSM_FORCE |
191 vc_sched.fill_rate = 0;
192 vc_sched.fill_rate2 = cpu_share; /* tokens accumulated per interval */
193 vc_sched.interval = vc_sched.interval2 = 1000; /* milliseconds */
194 vc_sched.tokens = 100; /* initial allocation of tokens */
195 vc_sched.tokens_min = 50; /* need this many tokens to run */
196 vc_sched.tokens_max = 100; /* max accumulated number of tokens */
198 if (cpu_share == (uint32_t)VC_LIM_KEEP)
199 vc_sched.set_mask &= ~(VC_VXSM_FILL_RATE|VC_VXSM_FILL_RATE2);
201 /* guaranteed CPU corresponds to SCHED_SHARE flag being cleared */
202 if (cpu_sched_flags & VS_SCHED_CPU_GUARANTEED) {
204 vc_sched.fill_rate = vc_sched.fill_rate2;
207 new_flags = VC_VXF_SCHED_SHARE;
209 VC_SYSCALL(vc_set_sched(ctx, &vc_sched));
211 vc_flags.mask = VC_VXF_SCHED_FLAGS;
212 vc_flags.flagword = new_flags | VC_VXF_SCHED_HARD;
213 VC_SYSCALL(vc_set_cflags(ctx, &vc_flags));
218 struct pl_resources {
220 unsigned long long *limit;
223 #define WHITESPACE(buffer,index,len) \
224 while(isspace((int)buffer[index])) \
225 if (index < len) index++; else goto out;
227 #define VSERVERCONF "/etc/vservers/"
229 pl_get_limits(char *context, struct sliver_resources *slr)
233 size_t len = strlen(VSERVERCONF) + strlen(context) + NULLBYTE_SIZE;
234 char *conf = (char *)malloc(len + strlen("rlimits/openfd.hard"));
235 struct pl_resources *r;
236 struct pl_resources sliver_list[] = {
237 {"sched/fill-rate2", &slr->vs_cpu},
238 {"sched/fill-rate", &slr->vs_cpuguaranteed},
240 {"rlimits/nproc.hard", &slr->vs_nproc.hard},
241 {"rlimits/nproc.soft", &slr->vs_nproc.soft},
242 {"rlimits/nproc.min", &slr->vs_nproc.min},
244 {"rlimits/rss.hard", &slr->vs_rss.hard},
245 {"rlimits/rss.soft", &slr->vs_rss.soft},
246 {"rlimits/rss.min", &slr->vs_rss.min},
248 {"rlimits/as.hard", &slr->vs_as.hard},
249 {"rlimits/as.soft", &slr->vs_as.soft},
250 {"rlimits/as.min", &slr->vs_as.min},
252 {"rlimits/openfd.hard", &slr->vs_openfd.hard},
253 {"rlimits/openfd.soft", &slr->vs_openfd.soft},
254 {"rlimits/openfd.min", &slr->vs_openfd.min},
256 {"whitelisted", &slr->vs_whitelisted},
258 {"bcapabilities", NULL},
262 sprintf(conf, "%s%s", VSERVERCONF, context);
264 slr->vs_cpu = VC_LIM_KEEP;
265 slr->vs_cpuguaranteed = 0;
267 slr->vs_rss.hard = VC_LIM_KEEP;
268 slr->vs_rss.soft = VC_LIM_KEEP;
269 slr->vs_rss.min = VC_LIM_KEEP;
271 slr->vs_as.hard = VC_LIM_KEEP;
272 slr->vs_as.soft = VC_LIM_KEEP;
273 slr->vs_as.min = VC_LIM_KEEP;
276 slr->vs_nproc.hard = VC_LIM_KEEP;
277 slr->vs_nproc.soft = VC_LIM_KEEP;
278 slr->vs_nproc.min = VC_LIM_KEEP;
280 slr->vs_openfd.hard = VC_LIM_KEEP;
281 slr->vs_openfd.soft = VC_LIM_KEEP;
282 slr->vs_openfd.min = VC_LIM_KEEP;
284 slr->vs_whitelisted = 1;
286 slr->vs_capabilities.bcaps = 0;
287 slr->vs_capabilities.bmask = 0;
288 slr->vs_capabilities.ccaps = 0;
289 slr->vs_capabilities.cmask = 0;
291 cwd = open(".", O_RDONLY);
293 perror("cannot get a handle on .");
296 if (chdir(conf) == -1) {
297 fprintf(stderr, "cannot chdir to ");
302 for (r = &sliver_list[0]; r->name; r++) {
304 fb = fopen(r->name, "r");
308 if (strcmp(r->name, "bcapabilities") == 0) {
310 struct vc_err_listparser err;
312 len = fread(buf, 1, sizeof(buf), fb);
313 for (i = 0; i < len; i++) {
317 vc_list2bcap(buf, len, &err, &slr->vs_capabilities);
320 if (fgets(buf, sizeof(buf), fb) != NULL && isdigit(*buf))
321 *r->limit = atoi(buf);
333 adjust_lim(struct vc_rlimit *vcr, struct rlimit *lim)
336 if (vcr->min != VC_LIM_KEEP) {
337 if (vcr->min > lim->rlim_cur) {
338 lim->rlim_cur = vcr->min;
341 if (vcr->min > lim->rlim_max) {
342 lim->rlim_max = vcr->min;
347 if (vcr->soft != VC_LIM_KEEP) {
348 switch (vcr->min != VC_LIM_KEEP) {
350 if (vcr->soft < vcr->min)
353 lim->rlim_cur = vcr->soft;
358 if (vcr->hard != VC_LIM_KEEP) {
359 switch (vcr->min != VC_LIM_KEEP) {
361 if (vcr->hard < vcr->min)
364 lim->rlim_cur = vcr->hard;
372 pl_set_limits(xid_t ctx, struct sliver_resources *slr)
374 struct rlimit lim; /* getrlimit values */
375 unsigned long long vs_cpu;
376 uint32_t cpu_sched_flags;
379 /* set memory limits */
380 getrlimit(RLIMIT_RSS,&lim);
381 if (adjust_lim(&slr->vs_rss, &lim)) {
382 setrlimit(RLIMIT_RSS, &lim);
383 if (vc_set_rlimit(ctx, RLIMIT_RSS, &slr->vs_rss))
385 PERROR("pl_setrlimit(%u, RLIMIT_RSS)", ctx);
390 /* set address space limits */
391 getrlimit(RLIMIT_AS,&lim);
392 if (adjust_lim(&slr->vs_as, &lim)) {
393 setrlimit(RLIMIT_AS, &lim);
394 if (vc_set_rlimit(ctx, RLIMIT_AS, &slr->vs_as))
396 PERROR("pl_setrlimit(%u, RLIMIT_AS)", ctx);
400 /* set nrpoc limit */
401 getrlimit(RLIMIT_NPROC,&lim);
402 if (adjust_lim(&slr->vs_nproc, &lim)) {
403 setrlimit(RLIMIT_NPROC, &lim);
404 if (vc_set_rlimit(ctx, RLIMIT_NPROC, &slr->vs_nproc))
406 PERROR("pl_setrlimit(%u, RLIMIT_NPROC)", ctx);
411 /* set openfd limit */
412 getrlimit(RLIMIT_NOFILE,&lim);
413 if (adjust_lim(&slr->vs_openfd, &lim)) {
414 setrlimit(RLIMIT_NOFILE, &lim);
415 if (vc_set_rlimit(ctx, RLIMIT_NOFILE, &slr->vs_openfd))
417 PERROR("pl_setrlimit(%u, RLIMIT_NOFILE)", ctx);
420 if (vc_set_rlimit(ctx, VC_VLIMIT_OPENFD, &slr->vs_openfd))
422 PERROR("pl_setrlimit(%u, VLIMIT_OPENFD)", ctx);
426 vs_cpu = slr->vs_cpu;
427 cpu_sched_flags = slr->vs_cpuguaranteed & VS_SCHED_CPU_GUARANTEED;
429 slr->vs_capabilities.bmask = vc_get_insecurebcaps();
430 if (vc_set_ccaps(ctx, &slr->vs_capabilities) < 0) {
431 PERROR("pl_setcaps(%u)", ctx);
439 if (pl_setsched(ctx, vs_cpu, cpu_sched_flags) < 0) {
440 PERROR("pl_setsched(%u)", ctx);