#include "macro.h"
#include "mainloop.h"
#include "namespace.h"
+#include "raw_syscalls.h"
#include "terminal.h"
#include "utils.h"
return ret;
}
-/**
- * This is based on raw_clone in systemd but adapted to our needs. This uses
- * copy on write semantics and doesn't pass a stack. CLONE_VM is tricky and
- * doesn't really matter to us so disallow it.
- *
- * The nice thing about this is that we get fork() behavior. That is
- * lxc_raw_clone() returns 0 in the child and the child pid in the parent.
- */
-pid_t lxc_raw_clone(unsigned long flags)
-{
-
- /* These flags don't interest at all so we don't jump through any hoopes
- * of retrieving them and passing them to the kernel.
- */
- errno = EINVAL;
- if ((flags & (CLONE_VM | CLONE_PARENT_SETTID | CLONE_CHILD_SETTID |
- CLONE_CHILD_CLEARTID | CLONE_SETTLS)))
- return -EINVAL;
-
-#if defined(__s390x__) || defined(__s390__) || defined(__CRIS__)
- /* On s390/s390x and cris the order of the first and second arguments
- * of the system call is reversed.
- */
- return (int)syscall(__NR_clone, NULL, flags | SIGCHLD);
-#elif defined(__sparc__) && defined(__arch64__)
- {
- /**
- * sparc64 always returns the other process id in %o0, and
- * a boolean flag whether this is the child or the parent in
- * %o1. Inline assembly is needed to get the flag returned
- * in %o1.
- */
- int in_child;
- int child_pid;
- asm volatile("mov %2, %%g1\n\t"
- "mov %3, %%o0\n\t"
- "mov 0 , %%o1\n\t"
- "t 0x6d\n\t"
- "mov %%o1, %0\n\t"
- "mov %%o0, %1"
- : "=r"(in_child), "=r"(child_pid)
- : "i"(__NR_clone), "r"(flags | SIGCHLD)
- : "%o1", "%o0", "%g1");
-
- if (in_child)
- return 0;
- else
- return child_pid;
- }
-#elif defined(__ia64__)
- /* On ia64 the stack and stack size are passed as separate arguments. */
- return (int)syscall(__NR_clone, flags | SIGCHLD, NULL, 0);
-#else
- return (int)syscall(__NR_clone, flags | SIGCHLD, NULL);
-#endif
-}
-
-pid_t lxc_raw_clone_cb(int (*fn)(void *), void *args, unsigned long flags)
-{
- pid_t pid;
-
- pid = lxc_raw_clone(flags);
- if (pid < 0)
- return -1;
-
- /* exit() is not thread-safe and might mess with the parent's signal
- * handlers and other stuff when exec() fails.
- */
- if (pid == 0)
- _exit(fn(args));
-
- return pid;
-}
-
/* Leave the user namespace at the first position in the array of structs so
* that we always attach to it first when iterating over the struct and using
* setns() to switch namespaces. This especially affects lxc_attach(): Suppose
* corresponding libc wrapper. glibc currently does not run pthread_atfork()
* handlers but does not guarantee that they are not. Other libcs might or
* might not run pthread_atfork() handlers. If you require guarantees please
- * refer to the lxc_raw_clone*() functions below.
+ * refer to the lxc_raw_clone*() functions in raw_syscalls.{c,h}.
*
* - should call lxc_raw_getpid():
* The child should use lxc_raw_getpid() to retrieve its pid.
*/
extern pid_t lxc_clone(int (*fn)(void *), void *arg, int flags);
-/**
- * lxc_raw_clone() - create a new process
- *
- * - fork() behavior:
- * This function returns 0 in the child and > 0 in the parent.
- *
- * - copy-on-write:
- * This function does not allocate a new stack and relies on copy-on-write
- * semantics.
- *
- * - supports subset of ClONE_* flags:
- * lxc_raw_clone() intentionally only supports a subset of the flags available
- * to the actual system call. Please refer to the implementation what flags
- * cannot be used. Also, please don't assume that just because a flag isn't
- * explicitly checked for as being unsupported that it is supported. If in
- * doubt or not sufficiently familiar with process creation in the kernel and
- * interactions with libcs this function should be used.
- *
- * - no pthread_atfork() handlers:
- * This function circumvents - as much as this this is possible - any libc
- * wrappers and thus does not run any pthread_atfork() handlers. Make sure
- * that this is safe to do in the context you are trying to call this
- * function.
- *
- * - must call lxc_raw_getpid():
- * The child must use lxc_raw_getpid() to retrieve its pid.
- */
-extern pid_t lxc_raw_clone(unsigned long flags);
-/**
- * lxc_raw_clone_cb() - create a new process
- *
- * - non-fork() behavior:
- * Function does return pid of the child or -1 on error. Pass in a callback
- * function via the "fn" argument that gets executed in the child process. The
- * "args" argument is passed to "fn".
- *
- * All other comments that apply to lxc_raw_clone() apply to lxc_raw_clone_cb()
- * as well.
- */
-extern pid_t lxc_raw_clone_cb(int (*fn)(void *), void *args,
- unsigned long flags);
-
extern int lxc_namespace_2_cloneflag(const char *namespace);
extern int lxc_namespace_2_ns_idx(const char *namespace);
extern int lxc_namespace_2_std_identifiers(char *namespaces);
#define _GNU_SOURCE 1
#endif
#include <errno.h>
+#include <sched.h>
+#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <sys/syscall.h>
#include <unistd.h>
#include "config.h"
+#include "macro.h"
+#include "raw_syscalls.h"
int lxc_raw_execveat(int dirfd, const char *pathname, char *const argv[],
char *const envp[], int flags)
return -1;
#endif
}
+
+/*
+ * This is based on raw_clone in systemd but adapted to our needs. This uses
+ * copy on write semantics and doesn't pass a stack. CLONE_VM is tricky and
+ * doesn't really matter to us so disallow it.
+ *
+ * The nice thing about this is that we get fork() behavior. That is
+ * lxc_raw_clone() returns 0 in the child and the child pid in the parent.
+ */
+pid_t lxc_raw_clone(unsigned long flags)
+{
+
+ /*
+ * These flags don't interest at all so we don't jump through any hoopes
+ * of retrieving them and passing them to the kernel.
+ */
+ errno = EINVAL;
+ if ((flags & (CLONE_VM | CLONE_PARENT_SETTID | CLONE_CHILD_SETTID |
+ CLONE_CHILD_CLEARTID | CLONE_SETTLS)))
+ return -EINVAL;
+
+#if defined(__s390x__) || defined(__s390__) || defined(__CRIS__)
+ /* On s390/s390x and cris the order of the first and second arguments
+ * of the system call is reversed.
+ */
+ return (int)syscall(__NR_clone, NULL, flags | SIGCHLD);
+#elif defined(__sparc__) && defined(__arch64__)
+ {
+ /*
+ * sparc64 always returns the other process id in %o0, and a
+ * boolean flag whether this is the child or the parent in %o1.
+ * Inline assembly is needed to get the flag returned in %o1.
+ */
+ int in_child;
+ int child_pid;
+ asm volatile("mov %2, %%g1\n\t"
+ "mov %3, %%o0\n\t"
+ "mov 0 , %%o1\n\t"
+ "t 0x6d\n\t"
+ "mov %%o1, %0\n\t"
+ "mov %%o0, %1"
+ : "=r"(in_child), "=r"(child_pid)
+ : "i"(__NR_clone), "r"(flags | SIGCHLD)
+ : "%o1", "%o0", "%g1");
+
+ if (in_child)
+ return 0;
+ else
+ return child_pid;
+ }
+#elif defined(__ia64__)
+ /* On ia64 the stack and stack size are passed as separate arguments. */
+ return (int)syscall(__NR_clone, flags | SIGCHLD, NULL, prctl_arg(0));
+#else
+ return (int)syscall(__NR_clone, flags | SIGCHLD, NULL);
+#endif
+}
+
+pid_t lxc_raw_clone_cb(int (*fn)(void *), void *args, unsigned long flags)
+{
+ pid_t pid;
+
+ pid = lxc_raw_clone(flags);
+ if (pid < 0)
+ return -1;
+
+ /*
+ * exit() is not thread-safe and might mess with the parent's signal
+ * handlers and other stuff when exec() fails.
+ */
+ if (pid == 0)
+ _exit(fn(args));
+
+ return pid;
+}
#ifndef _GNU_SOURCE
#define _GNU_SOURCE 1
#endif
+#include <sched.h>
#include <stdio.h>
#include <stdlib.h>
+#include <sys/syscall.h>
+
+/*
+ * lxc_raw_clone() - create a new process
+ *
+ * - fork() behavior:
+ * This function returns 0 in the child and > 0 in the parent.
+ *
+ * - copy-on-write:
+ * This function does not allocate a new stack and relies on copy-on-write
+ * semantics.
+ *
+ * - supports subset of ClONE_* flags:
+ * lxc_raw_clone() intentionally only supports a subset of the flags available
+ * to the actual system call. Please refer to the implementation what flags
+ * cannot be used. Also, please don't assume that just because a flag isn't
+ * explicitly checked for as being unsupported that it is supported. If in
+ * doubt or not sufficiently familiar with process creation in the kernel and
+ * interactions with libcs this function should be used.
+ *
+ * - no pthread_atfork() handlers:
+ * This function circumvents - as much as this this is possible - any libc
+ * wrappers and thus does not run any pthread_atfork() handlers. Make sure
+ * that this is safe to do in the context you are trying to call this
+ * function.
+ *
+ * - must call lxc_raw_getpid():
+ * The child must use lxc_raw_getpid() to retrieve its pid.
+ */
+extern pid_t lxc_raw_clone(unsigned long flags);
+
+/*
+ * lxc_raw_clone_cb() - create a new process
+ *
+ * - non-fork() behavior:
+ * Function does return pid of the child or -1 on error. Pass in a callback
+ * function via the "fn" argument that gets executed in the child process.
+ * The "args" argument is passed to "fn".
+ *
+ * All other comments that apply to lxc_raw_clone() apply to lxc_raw_clone_cb()
+ * as well.
+ */
+extern pid_t lxc_raw_clone_cb(int (*fn)(void *), void *args, unsigned long flags);
extern int lxc_raw_execveat(int dirfd, const char *pathname, char *const argv[],
char *const envp[], int flags);
#include "monitor.h"
#include "namespace.h"
#include "network.h"
+#include "raw_syscalls.h"
#include "start.h"
#include "storage/storage.h"
#include "storage/storage_utils.h"
#include "lxclock.h"
#include "namespace.h"
#include "parse.h"
+#include "raw_syscalls.h"
#include "syscall_wrappers.h"
#include "utils.h"
#include <sys/types.h>
#include "lxctest.h"
-#include "namespace.h"
+#include "raw_syscalls.h"
#include "utils.h"
int main(int argc, char *argv[])