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"
51 #ifndef VC_NXC_RAW_SOCKET
52 # define VC_NXC_RAW_SOCKET 0x00000200ull
54 #ifndef VC_NXC_RAW_SEND
55 # define VC_NXC_RAW_SEND 0x00000400ull
57 #ifndef VC_NXF_LBACK_ALLOW
58 # define VC_NXF_LBACK_ALLOW 0x00000400ull
62 create_context(xid_t ctx, uint64_t bcaps)
64 struct vc_ctx_caps vc_caps;
65 struct vc_net_flags vc_nf;
66 struct vc_net_caps vc_ncaps;
67 uint32_t unshare_mask;
69 /* Create network context */
70 if (vc_net_create(ctx) == VC_NOCTX) {
76 /* Make the network context persistent */
77 vc_nf.mask = vc_nf.flagword = VC_NXF_PERSISTENT | VC_NXF_LBACK_ALLOW;
78 if (vc_set_nflags(ctx, &vc_nf))
81 /* Give it raw sockets capabilities */
82 vc_ncaps.ncaps = vc_ncaps.cmask = VC_NXC_RAW_ICMP | VC_NXC_RAW_SOCKET;
83 if (vc_set_ncaps(ctx, &vc_ncaps))
87 /* Create tag context */
88 if (vc_tag_create(ctx) == VC_NOCTX)
94 * Create context info - this sets the STATE_SETUP and STATE_INIT flags.
96 if (vc_ctx_create(ctx, 0) == VC_NOCTX)
99 /* Unshare the net namespace if the slice if requested in the local slice configuration */
100 unshare_mask = get_space_flag(ctx);
101 if (unshare_mask != 0) {
102 sys_unshare(unshare_mask);
103 vc_set_namespace(ctx, unshare_mask);
106 /* Set capabilities - these don't take effect until SETUP flag is unset */
107 vc_caps.bcaps = bcaps;
108 vc_caps.bmask = ~0ULL; /* currently unused */
109 vc_caps.ccaps = 0; /* don't want any of these */
110 vc_caps.cmask = ~0ULL;
111 if (vc_set_ccaps(ctx, &vc_caps))
114 if (pl_setsched(ctx, 0, 1) < 0) {
115 PERROR("pl_setsched(%u)", ctx);
123 pl_setup_done(xid_t ctx)
125 struct vc_ctx_flags vc_flags;
127 /* unset SETUP flag - this allows other processes to migrate */
128 /* set the PERSISTENT flag - so the context doesn't vanish */
129 /* Don't clear the STATE_INIT flag, as that would make us the init task. */
130 vc_flags.mask = VC_VXF_STATE_SETUP|VC_VXF_PERSISTENT;
131 vc_flags.flagword = VC_VXF_PERSISTENT;
132 if (vc_set_cflags(ctx, &vc_flags))
138 #define RETRY_LIMIT 10
141 pl_chcontext(xid_t ctx, uint64_t bcaps, const struct sliver_resources *slr)
144 int net_migrated = 0;
146 if (pl_set_ulimits(slr) != 0)
151 struct vc_ctx_flags vc_flags;
153 if (vc_get_cflags(ctx, &vc_flags))
158 /* context doesn't exist - create it */
159 if (create_context(ctx, bcaps))
162 /* another process beat us in a race */
165 /* another process is creating - poll the SETUP flag */
170 /* created context and migrated to it i.e., we're done */
174 /* check the SETUP flag */
175 if (vc_flags.flagword & VC_VXF_STATE_SETUP)
177 /* context is still being setup - wait a while then retry */
178 if (retry_count++ >= RETRY_LIMIT)
187 /* context has been setup */
189 if (net_migrated || !vc_net_migrate(ctx))
191 /* Unshare the net namespace if the slice if requested in the local slice configuration */
192 unshare_mask = get_space_flag(ctx);
193 if (unshare_mask != 0) {
194 vc_enter_namespace(ctx, unshare_mask);
197 if (!vc_tag_migrate(ctx) && !vc_ctx_migrate(ctx, 0))
202 /* context disappeared - retry */
208 /* it's okay for a syscall to fail because the context doesn't exist */
209 #define VC_SYSCALL(x) \
213 return errno == ESRCH ? 0 : -1; \
218 pl_setsched(xid_t ctx, uint32_t cpu_min, uint32_t cpu_share)
220 struct vc_set_sched vc_sched;
221 struct vc_ctx_flags vc_flags;
223 vc_sched.set_mask = (VC_VXSM_FILL_RATE | VC_VXSM_INTERVAL | VC_VXSM_TOKENS |
224 VC_VXSM_TOKENS_MIN | VC_VXSM_TOKENS_MAX | VC_VXSM_MSEC |
225 VC_VXSM_FILL_RATE2 | VC_VXSM_INTERVAL2 | VC_VXSM_FORCE);
226 vc_sched.fill_rate = cpu_min; /* percent reserved */
227 vc_sched.interval = 100;
228 vc_sched.fill_rate2 = cpu_share; /* best-effort fair share of unreserved */
229 vc_sched.interval2 = 1000; /* milliseconds */
230 vc_sched.tokens = 100; /* initial allocation of tokens */
231 vc_sched.tokens_min = 50; /* need this many tokens to run */
232 vc_sched.tokens_max = 100; /* max accumulated number of tokens */
235 if (cpu_share == (uint32_t)VC_LIM_KEEP)
236 vc_sched.set_mask &= ~(VC_VXSM_FILL_RATE|VC_VXSM_FILL_RATE2);
238 vc_sched.set_mask |= VC_VXSM_IDLE_TIME;
241 VC_SYSCALL(vc_set_sched(ctx, &vc_sched));
243 vc_flags.mask = VC_VXF_SCHED_FLAGS;
244 vc_flags.flagword = VC_VXF_SCHED_HARD;
245 VC_SYSCALL(vc_set_cflags(ctx, &vc_flags));
255 struct pl_resources {
259 unsigned long long *limit;
260 unsigned long int *personality;
264 #define WHITESPACE(buffer,index,len) \
265 while(isspace((int)buffer[index])) \
266 if (index < len) index++; else goto out;
268 #define VSERVERCONF "/etc/vservers/"
271 pl_get_limits(const char *context, struct sliver_resources *slr)
275 size_t len = strlen(VSERVERCONF) + strlen(context) + NULLBYTE_SIZE;
276 char *conf = (char *)malloc(len + strlen("rlimits/openfd.hard"));
277 struct pl_resources *r;
278 struct pl_resources sliver_list[] = {
279 {"sched/fill-rate2", TYPE_LONG, &slr->vs_cpu},
281 {"rlimits/nproc.hard", TYPE_LONG, &slr->vs_nproc.hard},
282 {"rlimits/nproc.soft", TYPE_LONG, &slr->vs_nproc.soft},
283 {"rlimits/nproc.min", TYPE_LONG, &slr->vs_nproc.min},
285 {"rlimits/rss.hard", TYPE_LONG, &slr->vs_rss.hard},
286 {"rlimits/rss.soft", TYPE_LONG, &slr->vs_rss.soft},
287 {"rlimits/rss.min", TYPE_LONG, &slr->vs_rss.min},
289 {"rlimits/as.hard", TYPE_LONG, &slr->vs_as.hard},
290 {"rlimits/as.soft", TYPE_LONG, &slr->vs_as.soft},
291 {"rlimits/as.min", TYPE_LONG, &slr->vs_as.min},
293 {"rlimits/openfd.hard", TYPE_LONG, &slr->vs_openfd.hard},
294 {"rlimits/openfd.soft", TYPE_LONG, &slr->vs_openfd.soft},
295 {"rlimits/openfd.min", TYPE_LONG, &slr->vs_openfd.min},
297 {"personality", TYPE_PERS, &slr->personality},
302 sprintf(conf, "%s%s", VSERVERCONF, context);
304 slr->vs_rss.hard = VC_LIM_KEEP;
305 slr->vs_rss.soft = VC_LIM_KEEP;
306 slr->vs_rss.min = VC_LIM_KEEP;
308 slr->vs_as.hard = VC_LIM_KEEP;
309 slr->vs_as.soft = VC_LIM_KEEP;
310 slr->vs_as.min = VC_LIM_KEEP;
312 slr->vs_nproc.hard = VC_LIM_KEEP;
313 slr->vs_nproc.soft = VC_LIM_KEEP;
314 slr->vs_nproc.min = VC_LIM_KEEP;
316 slr->vs_openfd.hard = VC_LIM_KEEP;
317 slr->vs_openfd.soft = VC_LIM_KEEP;
318 slr->vs_openfd.min = VC_LIM_KEEP;
320 slr->personality = 0;
322 cwd = open(".", O_RDONLY);
324 perror("cannot get a handle on .");
327 if (chdir(conf) == -1) {
328 fprintf(stderr, "cannot chdir to ");
333 for (r = &sliver_list[0]; r->name; r++) {
335 fb = fopen(r->name, "r");
338 if (fgets(buf, sizeof(buf), fb) != NULL) {
340 /* remove trailing newline */
341 if (buf[len-1] == '\n') {
345 if ( (r->type == TYPE_LONG) && isdigit(*buf)) {
346 *r->limit = atoi(buf);
347 } else if ( (r->type == TYPE_PERS) && isalpha(*buf)) {
348 unsigned long int res;
349 res = vc_str2personalitytype(buf,len);
350 if (res != VC_BAD_PERSONALITY) {
351 *r->personality = res;
367 adjust_lim(const struct vc_rlimit *vcr, struct rlimit *lim)
370 if (vcr->min != VC_LIM_KEEP) {
371 if (vcr->min > lim->rlim_cur) {
372 lim->rlim_cur = vcr->min;
375 if (vcr->min > lim->rlim_max) {
376 lim->rlim_max = vcr->min;
381 if (vcr->soft != VC_LIM_KEEP) {
382 switch (vcr->min != VC_LIM_KEEP) {
384 if (vcr->soft < vcr->min)
387 lim->rlim_cur = vcr->soft;
392 if (vcr->hard != VC_LIM_KEEP) {
393 switch (vcr->min != VC_LIM_KEEP) {
395 if (vcr->hard < vcr->min)
398 lim->rlim_max = vcr->hard;
406 set_one_ulimit(int resource, const struct vc_rlimit *limit)
409 getrlimit(resource, &lim);
410 adjust_lim(limit, &lim);
411 setrlimit(resource, &lim);
415 set_personality(unsigned long int personality_arg)
417 if (personality_arg == 0)
419 if (personality(personality_arg) < 0) {
426 pl_set_ulimits(const struct sliver_resources *slr)
431 set_one_ulimit(RLIMIT_RSS, &slr->vs_rss);
432 set_one_ulimit(RLIMIT_AS, &slr->vs_as);
433 set_one_ulimit(RLIMIT_NPROC, &slr->vs_nproc);
434 set_one_ulimit(RLIMIT_NOFILE, &slr->vs_openfd);
435 return set_personality(slr->personality);