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 if (!vc_tag_migrate(ctx) && !vc_ctx_migrate(ctx, 0))
196 /* context disappeared - retry */
202 /* it's okay for a syscall to fail because the context doesn't exist */
203 #define VC_SYSCALL(x) \
207 return errno == ESRCH ? 0 : -1; \
212 pl_setsched(xid_t ctx, uint32_t cpu_min, uint32_t cpu_share)
214 struct vc_set_sched vc_sched;
215 struct vc_ctx_flags vc_flags;
217 vc_sched.set_mask = (VC_VXSM_FILL_RATE | VC_VXSM_INTERVAL | VC_VXSM_TOKENS |
218 VC_VXSM_TOKENS_MIN | VC_VXSM_TOKENS_MAX | VC_VXSM_MSEC |
219 VC_VXSM_FILL_RATE2 | VC_VXSM_INTERVAL2 | VC_VXSM_FORCE);
220 vc_sched.fill_rate = cpu_min; /* percent reserved */
221 vc_sched.interval = 100;
222 vc_sched.fill_rate2 = cpu_share; /* best-effort fair share of unreserved */
223 vc_sched.interval2 = 1000; /* milliseconds */
224 vc_sched.tokens = 100; /* initial allocation of tokens */
225 vc_sched.tokens_min = 50; /* need this many tokens to run */
226 vc_sched.tokens_max = 100; /* max accumulated number of tokens */
229 if (cpu_share == (uint32_t)VC_LIM_KEEP)
230 vc_sched.set_mask &= ~(VC_VXSM_FILL_RATE|VC_VXSM_FILL_RATE2);
232 vc_sched.set_mask |= VC_VXSM_IDLE_TIME;
235 VC_SYSCALL(vc_set_sched(ctx, &vc_sched));
237 vc_flags.mask = VC_VXF_SCHED_FLAGS;
238 vc_flags.flagword = VC_VXF_SCHED_HARD;
239 VC_SYSCALL(vc_set_cflags(ctx, &vc_flags));
249 struct pl_resources {
253 unsigned long long *limit;
254 unsigned long int *personality;
258 #define WHITESPACE(buffer,index,len) \
259 while(isspace((int)buffer[index])) \
260 if (index < len) index++; else goto out;
262 #define VSERVERCONF "/etc/vservers/"
265 pl_get_limits(const char *context, struct sliver_resources *slr)
269 size_t len = strlen(VSERVERCONF) + strlen(context) + NULLBYTE_SIZE;
270 char *conf = (char *)malloc(len + strlen("rlimits/openfd.hard"));
271 struct pl_resources *r;
272 struct pl_resources sliver_list[] = {
273 {"sched/fill-rate2", TYPE_LONG, &slr->vs_cpu},
275 {"rlimits/nproc.hard", TYPE_LONG, &slr->vs_nproc.hard},
276 {"rlimits/nproc.soft", TYPE_LONG, &slr->vs_nproc.soft},
277 {"rlimits/nproc.min", TYPE_LONG, &slr->vs_nproc.min},
279 {"rlimits/rss.hard", TYPE_LONG, &slr->vs_rss.hard},
280 {"rlimits/rss.soft", TYPE_LONG, &slr->vs_rss.soft},
281 {"rlimits/rss.min", TYPE_LONG, &slr->vs_rss.min},
283 {"rlimits/as.hard", TYPE_LONG, &slr->vs_as.hard},
284 {"rlimits/as.soft", TYPE_LONG, &slr->vs_as.soft},
285 {"rlimits/as.min", TYPE_LONG, &slr->vs_as.min},
287 {"rlimits/openfd.hard", TYPE_LONG, &slr->vs_openfd.hard},
288 {"rlimits/openfd.soft", TYPE_LONG, &slr->vs_openfd.soft},
289 {"rlimits/openfd.min", TYPE_LONG, &slr->vs_openfd.min},
291 {"personality", TYPE_PERS, &slr->personality},
296 sprintf(conf, "%s%s", VSERVERCONF, context);
298 slr->vs_rss.hard = VC_LIM_KEEP;
299 slr->vs_rss.soft = VC_LIM_KEEP;
300 slr->vs_rss.min = VC_LIM_KEEP;
302 slr->vs_as.hard = VC_LIM_KEEP;
303 slr->vs_as.soft = VC_LIM_KEEP;
304 slr->vs_as.min = VC_LIM_KEEP;
306 slr->vs_nproc.hard = VC_LIM_KEEP;
307 slr->vs_nproc.soft = VC_LIM_KEEP;
308 slr->vs_nproc.min = VC_LIM_KEEP;
310 slr->vs_openfd.hard = VC_LIM_KEEP;
311 slr->vs_openfd.soft = VC_LIM_KEEP;
312 slr->vs_openfd.min = VC_LIM_KEEP;
314 slr->personality = 0;
316 cwd = open(".", O_RDONLY);
318 perror("cannot get a handle on .");
321 if (chdir(conf) == -1) {
322 fprintf(stderr, "cannot chdir to ");
327 for (r = &sliver_list[0]; r->name; r++) {
329 fb = fopen(r->name, "r");
332 if (fgets(buf, sizeof(buf), fb) != NULL) {
334 /* remove trailing newline */
335 if (buf[len-1] == '\n') {
339 if ( (r->type == TYPE_LONG) && isdigit(*buf)) {
340 *r->limit = atoi(buf);
341 } else if ( (r->type == TYPE_PERS) && isalpha(*buf)) {
342 unsigned long int res;
343 res = vc_str2personalitytype(buf,len);
344 if (res != VC_BAD_PERSONALITY) {
345 *r->personality = res;
361 adjust_lim(const struct vc_rlimit *vcr, struct rlimit *lim)
364 if (vcr->min != VC_LIM_KEEP) {
365 if (vcr->min > lim->rlim_cur) {
366 lim->rlim_cur = vcr->min;
369 if (vcr->min > lim->rlim_max) {
370 lim->rlim_max = vcr->min;
375 if (vcr->soft != VC_LIM_KEEP) {
376 switch (vcr->min != VC_LIM_KEEP) {
378 if (vcr->soft < vcr->min)
381 lim->rlim_cur = vcr->soft;
386 if (vcr->hard != VC_LIM_KEEP) {
387 switch (vcr->min != VC_LIM_KEEP) {
389 if (vcr->hard < vcr->min)
392 lim->rlim_max = vcr->hard;
400 set_one_ulimit(int resource, const struct vc_rlimit *limit)
403 getrlimit(resource, &lim);
404 adjust_lim(limit, &lim);
405 setrlimit(resource, &lim);
409 set_personality(unsigned long int personality_arg)
411 if (personality_arg == 0)
413 if (personality(personality_arg) < 0) {
420 pl_set_ulimits(const struct sliver_resources *slr)
425 set_one_ulimit(RLIMIT_RSS, &slr->vs_rss);
426 set_one_ulimit(RLIMIT_AS, &slr->vs_as);
427 set_one_ulimit(RLIMIT_NPROC, &slr->vs_nproc);
428 set_one_ulimit(RLIMIT_NOFILE, &slr->vs_openfd);
429 return set_personality(slr->personality);