sanchayanmaity/minix
1
0
Fork 0
Code Issues Pull requests Packages Projects Releases Wiki Activity
minix/servers/vfs/exec.c

754 lines
23 KiB
C
Raw Normal View History

New interface between PM and FS.
2006-05-11 16:57:23 +02:00
/* This file handles the EXEC system call. It performs the work as follows:
* - see if the permissions allow the file to be executed
* - read the header and extract the sizes
* - fetch the initial args and environment from the user space
* - allocate the memory for the new process
* - copy the initial stack from PM to the process
* - read in the text and data segments and copy to the process
* - take care of setuid and setgid bits
* - fix up 'mproc' table
* - tell kernel about EXEC
* - save offset to initial argc (for ps)
*
* The entry points into this file are:
* pm_exec: perform the EXEC system call
*/
#include "fs.h"
#include <sys/stat.h>
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
#include <sys/mman.h>
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
#include <minix/callnr.h>
#include <minix/endpoint.h>
#include <minix/com.h>
First cut at 64-bit file offsets in block devices for mkfs/fsck.
2006-11-27 15:21:43 +01:00
#include <minix/u64.h>
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
#include <signal.h>
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
#include <stdlib.h>
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
#include <string.h>
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
#include <dirent.h>
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
#include <sys/param.h>
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
#include "fproc.h"
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
#include "path.h"
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
#include "param.h"
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
#include "vnode.h"
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
#include "file.h"
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
#include <minix/vfsif.h>
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
#include <machine/vmparam.h>
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
#include <assert.h>
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
#include <fcntl.h>
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
#define _KERNEL /* for ELF_AUX_ENTRIES */
#include <libexec.h>
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
/* fields only used by elf and in VFS */
struct vfs_exec_info {
struct exec_info args; /* libexec exec args */
struct vnode *vp; /* Exec file's vnode */
struct vmnt *vmp; /* Exec file's vmnt */
struct stat sb; /* Exec file's stat structure */
int userflags; /* exec() flags from userland */
int is_dyn; /* Dynamically linked executable */
int elf_main_fd; /* Dyn: FD of main program execuatble */
char execname[PATH_MAX]; /* Full executable invocation */
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
int vmfd;
int vmfd_used;
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
};
retire PUBLIC, PRIVATE and FORWARD
2012-03-25 20:25:53 +02:00
static void lock_exec(void);
static void unlock_exec(void);
VFS: make m_in job local By making m_in job local (i.e., each job has its own copy of m_in instead of refering to the global m_in) we don't have to store and restore m_in on every thread yield. This reduces overhead. Moreover, remove the assumption that m_in is preserved. Do_XXX functions have to copy the system call parameters as soon as possible and only pass those copies to other functions. Furthermore, this patch cleans up some code and uses better types in a lot of places.
2012-04-13 14:50:38 +02:00
static int patch_stack(struct vnode *vp, char stack[ARG_MAX],
size_t *stk_bytes, char path[PATH_MAX]);
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
static int is_script(struct vfs_exec_info *execi);
VFS: make m_in job local By making m_in job local (i.e., each job has its own copy of m_in instead of refering to the global m_in) we don't have to store and restore m_in on every thread yield. This reduces overhead. Moreover, remove the assumption that m_in is preserved. Do_XXX functions have to copy the system call parameters as soon as possible and only pass those copies to other functions. Furthermore, this patch cleans up some code and uses better types in a lot of places.
2012-04-13 14:50:38 +02:00
static int insert_arg(char stack[ARG_MAX], size_t *stk_bytes, char *arg,
int replace);
retire PUBLIC, PRIVATE and FORWARD
2012-03-25 20:25:53 +02:00
static void clo_exec(struct fproc *rfp);
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
static int stack_prepare_elf(struct vfs_exec_info *execi,
AT_SUN_EXECNAME support . vfs: pass execname in aux vectors . ld.elf_so: use this to expand $ORIGIN . this requires the executable to reserve more space at exec() calling time
2012-04-18 16:32:38 +02:00
char *curstack, size_t *frame_len, vir_bytes *vsp, int *extrabase);
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
static int map_header(struct vfs_exec_info *execi);
static int read_seg(struct exec_info *execi, off_t off, off_t seg_addr, size_t seg_bytes);
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
#define PTRSIZE sizeof(char *) /* Size of pointers in argv[] and envp[]. */
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
/* Array of loaders for different object file formats */
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
typedef int (*exechook_t)(struct vfs_exec_info *execpackage);
typedef int (*stackhook_t)(struct vfs_exec_info *execi, char *curstack,
AT_SUN_EXECNAME support . vfs: pass execname in aux vectors . ld.elf_so: use this to expand $ORIGIN . this requires the executable to reserve more space at exec() calling time
2012-04-18 16:32:38 +02:00
size_t *frame_len, vir_bytes *, int *extrabase);
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
struct exec_loaders {
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
libexec_exec_loadfunc_t load_object; /* load executable into memory */
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
stackhook_t setup_stack; /* prepare stack before argc and argv push */
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
};
retire PUBLIC, PRIVATE and FORWARD
2012-03-25 20:25:53 +02:00
static const struct exec_loaders exec_loaders[] = {
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
{ libexec_load_elf, stack_prepare_elf },
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
{ NULL, NULL }
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
};
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
/*===========================================================================*
* lock_exec *
*===========================================================================*/
retire PUBLIC, PRIVATE and FORWARD
2012-03-25 20:25:53 +02:00
static void lock_exec(void)
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
{
struct fproc *org_fp;
struct worker_thread *org_self;
/* First try to get it right off the bat */
if (mutex_trylock(&exec_lock) == 0)
return;
org_fp = fp;
org_self = self;
if (mutex_lock(&exec_lock) != 0)
panic("Could not obtain lock on exec");
fp = org_fp;
self = org_self;
}
/*===========================================================================*
* unlock_exec *
*===========================================================================*/
retire PUBLIC, PRIVATE and FORWARD
2012-03-25 20:25:53 +02:00
static void unlock_exec(void)
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
{
if (mutex_unlock(&exec_lock) != 0)
panic("Could not release lock on exec");
}
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
/*===========================================================================*
* get_read_vp *
*===========================================================================*/
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
static int get_read_vp(struct vfs_exec_info *execi,
char *fullpath, int copyprogname, int sugid, struct lookup *resolve, struct fproc *fp)
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
{
/* Make the executable that we want to exec() into the binary pointed
* to by 'fullpath.' This function fills in necessary details in the execi
* structure, such as opened vnode. It unlocks and releases the vnode if
* it was already there. This makes it easy to change the executable
* during the exec(), which is often necessary, by calling this function
* more than once. This is specifically necessary when we discover the
* executable is actually a script or a dynamically linked executable.
*/
int r;
/* Caller wants to switch vp to the file in 'fullpath.'
* unlock and put it first if there is any there.
*/
if(execi->vp) {
unlock_vnode(execi->vp);
put_vnode(execi->vp);
execi->vp = NULL;
}
/* Remember/overwrite the executable name if requested. */
if(copyprogname) {
char *cp = strrchr(fullpath, '/');
if(cp) cp++;
else cp = fullpath;
VFS: Use safe string copy functions
2012-07-13 18:08:06 +02:00
strlcpy(execi->args.progname, cp, sizeof(execi->args.progname));
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
execi->args.progname[sizeof(execi->args.progname)-1] = '\0';
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
}
/* Open executable */
if ((execi->vp = eat_path(resolve, fp)) == NULL)
return err_code;
unlock_vmnt(execi->vmp);
VFS: use S_IS* macros
2012-04-25 14:44:42 +02:00
if (!S_ISREG(execi->vp->v_mode))
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
return ENOEXEC;
else if ((r = forbidden(fp, execi->vp, X_BIT)) != OK)
return r;
else
r = req_stat(execi->vp->v_fs_e, execi->vp->v_inode_nr,
Spring cleanup Remove old versions of system calls and system calls that don't have a libc api interface anymore (dup, dup2, creat). VFS still contains support for old system call numbers for the new stat system calls (i.e., 65, 66, 67) to keep supporting old binaries built for MINIX 3.2.1 (prior to the release). Change-Id: I721779b58a50c7eeae20669de24658d55d69b25b
2013-02-28 15:44:23 +01:00
VFS_PROC_NR, (vir_bytes) &(execi->sb));
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
if (r != OK) return r;
/* If caller wants us to, honour suid/guid mode bits. */
if (sugid) {
/* Deal with setuid/setgid executables */
if (execi->vp->v_mode & I_SET_UID_BIT) {
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
execi->args.new_uid = execi->vp->v_uid;
execi->args.allow_setuid = 1;
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
}
if (execi->vp->v_mode & I_SET_GID_BIT) {
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
execi->args.new_gid = execi->vp->v_gid;
execi->args.allow_setuid = 1;
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
}
}
/* Read in first chunk of file. */
if((r=map_header(execi)) != OK)
return r;
return OK;
}
#define FAILCHECK(expr) if((r=(expr)) != OK) { goto pm_execfinal; } while(0)
#define Get_read_vp(e,f,p,s,rs,fp) do { \
r=get_read_vp(&e,f,p,s,rs,fp); if(r != OK) { FAILCHECK(r); } \
} while(0)
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
static int vfs_memmap(struct exec_info *execi,
vir_bytes vaddr, vir_bytes len, vir_bytes foffset, u16_t clearend,
int protflags)
{
struct vfs_exec_info *vi = (struct vfs_exec_info *) execi->opaque;
struct vnode *vp = ((struct vfs_exec_info *) execi->opaque)->vp;
int r;
u16_t flags = 0;
if(protflags & PROT_WRITE)
flags |= MVM_WRITABLE;
r = minix_vfs_mmap(execi->proc_e, foffset, len,
vp->v_dev, vp->v_inode_nr, vi->vmfd, vaddr, clearend, flags);
if(r == OK) {
vi->vmfd_used = 1;
}
return r;
}
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
/*===========================================================================*
* pm_exec *
*===========================================================================*/
VFS: make m_in job local By making m_in job local (i.e., each job has its own copy of m_in instead of refering to the global m_in) we don't have to store and restore m_in on every thread yield. This reduces overhead. Moreover, remove the assumption that m_in is preserved. Do_XXX functions have to copy the system call parameters as soon as possible and only pass those copies to other functions. Furthermore, this patch cleans up some code and uses better types in a lot of places.
2012-04-13 14:50:38 +02:00
int pm_exec(endpoint_t proc_e, vir_bytes path, size_t path_len,
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
vir_bytes frame, size_t frame_len, vir_bytes *pc,
vir_bytes *newsp, int user_exec_flags)
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
{
/* Perform the execve(name, argv, envp) call. The user library builds a
* complete stack image, including pointers, args, environ, etc. The stack
More cleanup to remove MM and FS references
2010-06-10 16:04:46 +02:00
* is copied to a buffer inside VFS, and then to the new core image.
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
*/
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
int r, slot;
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
vir_bytes vsp;
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
struct fproc *rfp;
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
int extrabase = 0;
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
static char mbuf[ARG_MAX]; /* buffer for stack and zeroes */
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
struct vfs_exec_info execi;
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
int i;
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
static char fullpath[PATH_MAX],
elf_interpreter[PATH_MAX],
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
firstexec[PATH_MAX],
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
finalexec[PATH_MAX];
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
struct lookup resolve;
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
struct fproc *vmfp = &fproc[VM_PROC_NR];
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
stackhook_t makestack = NULL;
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
static int n;
n++;
struct filp *newfilp = NULL;
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
lock_exec();
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
lock_proc(vmfp, 0);
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
/* unset execi values are 0. */
memset(&execi, 0, sizeof(execi));
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
execi.vmfd = -1;
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
/* passed from exec() libc code */
execi.userflags = user_exec_flags;
No more intel/minix segments. This commit removes all traces of Minix segments (the text/data/stack memory map abstraction in the kernel) and significance of Intel segments (hardware segments like CS, DS that add offsets to all addressing before page table translation). This ultimately simplifies the memory layout and addressing and makes the same layout possible on non-Intel architectures. There are only two types of addresses in the world now: virtual and physical; even the kernel and processes have the same virtual address space. Kernel and user processes can be distinguished at a glance as processes won't use 0xF0000000 and above. No static pre-allocated memory sizes exist any more. Changes to booting: . The pre_init.c leaves the kernel and modules exactly as they were left by the bootloader in physical memory . The kernel starts running using physical addressing, loaded at a fixed location given in its linker script by the bootloader. All code and data in this phase are linked to this fixed low location. . It makes a bootstrap pagetable to map itself to a fixed high location (also in linker script) and jumps to the high address. All code and data then use this high addressing. . All code/data symbols linked at the low addresses is prefixed by an objcopy step with __k_unpaged_*, so that that code cannot reference highly-linked symbols (which aren't valid yet) or vice versa (symbols that aren't valid any more). . The two addressing modes are separated in the linker script by collecting the unpaged_*.o objects and linking them with low addresses, and linking the rest high. Some objects are linked twice, once low and once high. . The bootstrap phase passes a lot of information (e.g. free memory list, physical location of the modules, etc.) using the kinfo struct. . After this bootstrap the low-linked part is freed. . The kernel maps in VM into the bootstrap page table so that VM can begin executing. Its first job is to make page tables for all other boot processes. So VM runs before RS, and RS gets a fully dynamic, VM-managed address space. VM gets its privilege info from RS as usual but that happens after RS starts running. . Both the kernel loading VM and VM organizing boot processes happen using the libexec logic. This removes the last reason for VM to still know much about exec() and vm/exec.c is gone. Further Implementation: . All segments are based at 0 and have a 4 GB limit. . The kernel is mapped in at the top of the virtual address space so as not to constrain the user processes. . Processes do not use segments from the LDT at all; there are no segments in the LDT any more, so no LLDT is needed. . The Minix segments T/D/S are gone and so none of the user-space or in-kernel copy functions use them. The copy functions use a process endpoint of NONE to realize it's a physical address, virtual otherwise. . The umap call only makes sense to translate a virtual address to a physical address now. . Segments-related calls like newmap and alloc_segments are gone. . All segments-related translation in VM is gone (vir2map etc). . Initialization in VM is simpler as no moving around is necessary. . VM and all other boot processes can be linked wherever they wish and will be mapped in at the right location by the kernel and VM respectively. Other changes: . The multiboot code is less special: it does not use mb_print for its diagnostics any more but uses printf() as normal, saving the output into the diagnostics buffer, only printing to the screen using the direct print functions if a panic() occurs. . The multiboot code uses the flexible 'free memory map list' style to receive the list of free memory if available. . The kernel determines the memory layout of the processes to a degree: it tells VM where the kernel starts and ends and where the kernel wants the top of the process to be. VM then uses this entire range, i.e. the stack is right at the top, and mmap()ped bits of memory are placed below that downwards, and the break grows upwards. Other Consequences: . Every process gets its own page table as address spaces can't be separated any more by segments. . As all segments are 0-based, there is no distinction between virtual and linear addresses, nor between userspace and kernel addresses. . Less work is done when context switching, leading to a net performance increase. (8% faster on my machine for 'make servers'.) . The layout and configuration of the GDT makes sysenter and syscall possible.
2012-05-07 16:03:35 +02:00
execi.args.stack_high = kinfo.user_sp;
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
execi.args.stack_size = DEFAULT_STACK_LIMIT;
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
okendpt(proc_e, &slot);
rfp = fp = &fproc[slot];
Implement getrusage Implement getrusage. These fields of struct rusage are not supported and always set to zero at this time long ru_nswap; /* swaps */ long ru_inblock; /* block input operations */ long ru_oublock; /* block output operations */ long ru_msgsnd; /* messages sent */ long ru_msgrcv; /* messages received */ long ru_nvcsw; /* voluntary context switches */ long ru_nivcsw; /* involuntary context switches */ test75.c is the unit test for this new function Change-Id: I3f1eb69de1fce90d087d76773b09021fc6106539
2013-06-25 14:41:01 +02:00
rfp->text_size = 0;
rfp->data_size = 0;
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
lookup_init(&resolve, fullpath, PATH_NOFLAGS, &execi.vmp, &execi.vp);
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
resolve.l_vmnt_lock = VMNT_READ;
resolve.l_vnode_lock = VNODE_READ;
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
/* Fetch the stack from the user before destroying the old core image. */
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
if (frame_len > ARG_MAX)
FAILCHECK(ENOMEM); /* stack too big */
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
r = sys_datacopy(proc_e, (vir_bytes) frame, SELF, (vir_bytes) mbuf,
VFS: make m_in job local By making m_in job local (i.e., each job has its own copy of m_in instead of refering to the global m_in) we don't have to store and restore m_in on every thread yield. This reduces overhead. Moreover, remove the assumption that m_in is preserved. Do_XXX functions have to copy the system call parameters as soon as possible and only pass those copies to other functions. Furthermore, this patch cleans up some code and uses better types in a lot of places.
2012-04-13 14:50:38 +02:00
(size_t) frame_len);
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
if (r != OK) { /* can't fetch stack (e.g. bad virtual addr) */
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
printf("VFS: pm_exec: sys_datacopy failed\n");
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
FAILCHECK(r);
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
}
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
/* The default is to keep the original user and group IDs */
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
execi.args.new_uid = rfp->fp_effuid;
execi.args.new_gid = rfp->fp_effgid;
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
/* Get the exec file name. */
FAILCHECK(fetch_name(path, path_len, fullpath));
VFS: Use safe string copy functions
2012-07-13 18:08:06 +02:00
strlcpy(finalexec, fullpath, PATH_MAX);
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
strlcpy(firstexec, fullpath, PATH_MAX);
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
/* Get_read_vp will return an opened vn in execi.
* if necessary it releases the existing vp so we can
* switch after we find out what's inside the file.
* It reads the start of the file.
*/
Get_read_vp(execi, fullpath, 1, 1, &resolve, fp);
/* If this is a script (i.e. has a #!/interpreter line),
* retrieve the name of the interpreter and open that
* executable instead.
*/
if(is_script(&execi)) {
/* patch_stack will add interpreter name and
* args to stack and retrieve the new binary
* name into fullpath.
*/
VFS: Use safe string copy functions
2012-07-13 18:08:06 +02:00
FAILCHECK(fetch_name(path, path_len, fullpath));
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
FAILCHECK(patch_stack(execi.vp, mbuf, &frame_len, fullpath));
VFS: Use safe string copy functions
2012-07-13 18:08:06 +02:00
strlcpy(finalexec, fullpath, PATH_MAX);
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
strlcpy(firstexec, fullpath, PATH_MAX);
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
Get_read_vp(execi, fullpath, 1, 0, &resolve, fp);
}
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
/* If this is a dynamically linked executable, retrieve
* the name of that interpreter in elf_interpreter and open that
* executable instead. But open the current executable in an
* fd for the current process.
*/
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
if(elf_has_interpreter(execi.args.hdr, execi.args.hdr_len,
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
elf_interpreter, sizeof(elf_interpreter))) {
/* Switch the executable vnode to the interpreter */
execi.is_dyn = 1;
/* The interpreter (loader) needs an fd to the main program,
* which is currently in finalexec
*/
if((r = execi.elf_main_fd = common_open(finalexec, O_RDONLY, 0)) < 0) {
printf("VFS: exec: dynamic: open main exec failed %s (%d)\n",
fullpath, r);
FAILCHECK(r);
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
}
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
libexec: add load_offset feature, used for ld.so . ld.so is linked at 0 but it can relocate itself; we wish to load ld.so higher though to trap NULL dereferences. if we know we have to execute ld.so, vfs tells libexec to put it higher.
2012-08-10 18:27:23 +02:00
/* ld.so is linked at 0, but it can relocate itself; we
* want it higher to trap NULL pointer dereferences.
*/
execi.args.load_offset = 0x10000;
AT_SUN_EXECNAME support . vfs: pass execname in aux vectors . ld.elf_so: use this to expand $ORIGIN . this requires the executable to reserve more space at exec() calling time
2012-04-18 16:32:38 +02:00
/* Remember it */
VFS: Use safe string copy functions
2012-07-13 18:08:06 +02:00
strlcpy(execi.execname, finalexec, PATH_MAX);
AT_SUN_EXECNAME support . vfs: pass execname in aux vectors . ld.elf_so: use this to expand $ORIGIN . this requires the executable to reserve more space at exec() calling time
2012-04-18 16:32:38 +02:00
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
/* The executable we need to execute first (loader)
* is in elf_interpreter, and has to be in fullpath to
* be looked up
*/
VFS: Use safe string copy functions
2012-07-13 18:08:06 +02:00
strlcpy(fullpath, elf_interpreter, PATH_MAX);
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
strlcpy(firstexec, elf_interpreter, PATH_MAX);
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
Get_read_vp(execi, fullpath, 0, 0, &resolve, fp);
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
}
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
/* We also want an FD for VM to mmap() the process in if possible. */
{
struct vnode *vp = execi.vp;
assert(vp);
if(vp->v_vmnt->m_haspeek && major(vp->v_dev) != MEMORY_MAJOR) {
int newfd = -1;
if(get_fd(vmfp, 0, R_BIT, &newfd, &newfilp) == OK) {
assert(newfd >= 0 && newfd < OPEN_MAX);
assert(!vmfp->fp_filp[newfd]);
newfilp->filp_count = 1;
newfilp->filp_vno = vp;
newfilp->filp_flags = O_RDONLY;
FD_SET(newfd, &vmfp->fp_filp_inuse);
vmfp->fp_filp[newfd] = newfilp;
/* dup_vnode(vp); */
execi.vmfd = newfd;
execi.args.memmap = vfs_memmap;
}
}
}
further libexec generalization . new mode for sys_memset: include process so memset can be done in physical or virtual address space. . add a mode to mmap() that lets a process allocate uninitialized memory. . this allows an exec()er (RS, VFS, etc.) to request uninitialized memory from VM and selectively clear the ranges that don't come from a file, leaving no uninitialized memory left for the process to see. . use callbacks for clearing the process, clearing memory in the process, and copying into the process; so that the libexec code can be used from rs, vfs, and in the future, kernel (to load vm) and vm (to load boot-time processes)
2012-06-06 19:05:28 +02:00
/* callback functions and data */
execi.args.copymem = read_seg;
execi.args.clearproc = libexec_clearproc_vm_procctl;
execi.args.clearmem = libexec_clear_sys_memset;
libexec: mmap support, prealloc variants In libexec, split the memory allocation method into cleared and non-cleared. Cleared gives zeroed memory, non-cleared gives 'junk' memory (that will be overwritten anyway, and so needn't be cleared) that is faster to get. Also introduce the 'memmap' method that can be used, if available, to map code and data from executables into a process using the third-party mmap() mode. Change-Id: I26694fd3c21deb8b97e01ed675dfc14719b0672b
2013-03-16 04:46:37 +01:00
execi.args.allocmem_prealloc_cleared = libexec_alloc_mmap_prealloc_cleared;
execi.args.allocmem_prealloc_junk = libexec_alloc_mmap_prealloc_junk;
further libexec generalization . new mode for sys_memset: include process so memset can be done in physical or virtual address space. . add a mode to mmap() that lets a process allocate uninitialized memory. . this allows an exec()er (RS, VFS, etc.) to request uninitialized memory from VM and selectively clear the ranges that don't come from a file, leaving no uninitialized memory left for the process to see. . use callbacks for clearing the process, clearing memory in the process, and copying into the process; so that the libexec code can be used from rs, vfs, and in the future, kernel (to load vm) and vm (to load boot-time processes)
2012-06-06 19:05:28 +02:00
execi.args.allocmem_ondemand = libexec_alloc_mmap_ondemand;
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
execi.args.opaque = &execi;
further libexec generalization . new mode for sys_memset: include process so memset can be done in physical or virtual address space. . add a mode to mmap() that lets a process allocate uninitialized memory. . this allows an exec()er (RS, VFS, etc.) to request uninitialized memory from VM and selectively clear the ranges that don't come from a file, leaving no uninitialized memory left for the process to see. . use callbacks for clearing the process, clearing memory in the process, and copying into the process; so that the libexec code can be used from rs, vfs, and in the future, kernel (to load vm) and vm (to load boot-time processes)
2012-06-06 19:05:28 +02:00
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
execi.args.proc_e = proc_e;
execi.args.frame_len = frame_len;
libexec: detect short files if an exec() fails partway through reading in the sections, the target process is already gone and a defunct process remains. sanity checking the binary beforehand helps that. test10 mutilates binaries and exec()s them on purpose; making an exec() fail cleanly in such cases seems like acceptable behaviour. fixes test10 on ARM. Change-Id: I1ed9bb200ce469d4d349073cadccad5503b2fcb0
2013-02-04 01:49:48 +01:00
execi.args.filesize = execi.vp->v_size;
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
for (i = 0; exec_loaders[i].load_object != NULL; i++) {
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
r = (*exec_loaders[i].load_object)(&execi.args);
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
/* Loaded successfully, so no need to try other loaders */
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
if (r == OK) { makestack = exec_loaders[i].setup_stack; break; }
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
}
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
FAILCHECK(r);
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
further libexec generalization . new mode for sys_memset: include process so memset can be done in physical or virtual address space. . add a mode to mmap() that lets a process allocate uninitialized memory. . this allows an exec()er (RS, VFS, etc.) to request uninitialized memory from VM and selectively clear the ranges that don't come from a file, leaving no uninitialized memory left for the process to see. . use callbacks for clearing the process, clearing memory in the process, and copying into the process; so that the libexec code can be used from rs, vfs, and in the future, kernel (to load vm) and vm (to load boot-time processes)
2012-06-06 19:05:28 +02:00
/* Inform PM */
FAILCHECK(libexec_pm_newexec(proc_e, &execi.args));
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
/* Save off PC */
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
*pc = execi.args.pc;
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
/* call a stack-setup function if this executable type wants it */
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
vsp = execi.args.stack_high - frame_len;
AT_SUN_EXECNAME support . vfs: pass execname in aux vectors . ld.elf_so: use this to expand $ORIGIN . this requires the executable to reserve more space at exec() calling time
2012-04-18 16:32:38 +02:00
if(makestack) FAILCHECK(makestack(&execi, mbuf, &frame_len, &vsp, &extrabase));
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
/* Patch up stack and copy it from VFS to new core image. */
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
libexec_patch_ptr(mbuf, vsp + extrabase);
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
FAILCHECK(sys_datacopy(SELF, (vir_bytes) mbuf, proc_e, (vir_bytes) vsp,
(phys_bytes)frame_len));
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
/* Return new stack pointer to caller */
*newsp = vsp;
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
clo_exec(rfp);
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
if (execi.args.allow_setuid) {
Implement issetugid syscall Implement issetugid syscall and provide a test. This gets rid of the scary "Unsecure. Implement me" warning during compilation.
2011-11-28 11:03:43 +01:00
/* If after loading the image we're still allowed to run with
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
* setuid or setgid, change credentials now */
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
rfp->fp_effuid = execi.args.new_uid;
rfp->fp_effgid = execi.args.new_gid;
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
}
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
/* Remember the new name of the process */
VFS: Use safe string copy functions
2012-07-13 18:08:06 +02:00
strlcpy(rfp->fp_name, execi.args.progname, PROC_NAME_LEN);
Implement getrusage Implement getrusage. These fields of struct rusage are not supported and always set to zero at this time long ru_nswap; /* swaps */ long ru_inblock; /* block input operations */ long ru_oublock; /* block output operations */ long ru_msgsnd; /* messages sent */ long ru_msgrcv; /* messages received */ long ru_nvcsw; /* voluntary context switches */ long ru_nivcsw; /* involuntary context switches */ test75.c is the unit test for this new function Change-Id: I3f1eb69de1fce90d087d76773b09021fc6106539
2013-06-25 14:41:01 +02:00
rfp->text_size = execi.args.text_size;
rfp->data_size = execi.args.data_size;
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
pm_execfinal:
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
if(newfilp) unlock_filp(newfilp);
else if (execi.vp != NULL) {
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
unlock_vnode(execi.vp);
put_vnode(execi.vp);
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
}
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
if(execi.vmfd >= 0 && !execi.vmfd_used) {
if(OK != close_fd(vmfp, execi.vmfd)) {
printf("VFS: unexpected close fail of vm fd\n");
}
}
unlock_proc(vmfp);
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
unlock_exec();
vfs: mmap support . libc: add vfs_mmap, a way for vfs to initiate mmap()s. This is a good special case to have as vfs is a slightly different client from regular user processes. It doesn't do it for itself, and has the dev & inode info already so the callback to VFS for the lookup isn't necessary. So it has different info to have to give to VM. . libc: also add minix_mmap64() that accepts a 64-bit offset, even though our off_t is still 32 bit now. . On exec() time, try to mmap() in the executable if available. (It is not yet available in this commit.) . To support mmap(), add do_vm_call that allows VM to lookup (to ino+dev), do i/o from and close FD's on behalf of other processes. Change-Id: I831551e45a6781c74313c450eb9c967a68505932
2013-05-07 14:41:07 +02:00
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
return(r);
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
}
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
static int stack_prepare_elf(struct vfs_exec_info *execi, char *frame, size_t *framelen,
AT_SUN_EXECNAME support . vfs: pass execname in aux vectors . ld.elf_so: use this to expand $ORIGIN . this requires the executable to reserve more space at exec() calling time
2012-04-18 16:32:38 +02:00
vir_bytes *newsp, int *extrabase)
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
{
AT_SUN_EXECNAME support . vfs: pass execname in aux vectors . ld.elf_so: use this to expand $ORIGIN . this requires the executable to reserve more space at exec() calling time
2012-04-18 16:32:38 +02:00
AuxInfo *a, *term;
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
Elf_Ehdr *elf_header;
int nulls;
char **mysp = (char **) frame,
**mysp_end = (char **) ((char *)frame + *framelen);
if(!execi->is_dyn)
return OK;
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
assert(execi->args.hdr_len >= sizeof(*elf_header));
elf_header = (Elf_Ehdr *) execi->args.hdr;
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
/* exec() promises stack space. Now find it. */
mysp++; /* skip argc */
/* find a terminating NULL entry twice: one for argv[], one for envp[]. */
for(nulls = 0; nulls < 2; nulls++) {
assert(mysp < mysp_end);
while(*mysp && mysp < mysp_end) mysp++; /* find terminating NULL */
if(mysp >= mysp_end) {
printf("VFS: malformed stack for exec()\n");
return ENOEXEC;
}
assert(!*mysp);
mysp++;
}
/* Userland provides a fully filled stack frame, with argc, argv, envp
* and then all the argv and envp strings; consistent with ELF ABI, except
* for a list of Aux vectors that should be between envp points and the
* start of the strings.
*
* It would take some very unpleasant hackery to insert the aux vectors before
* the strings, and correct all the pointers, so the exec code in libc makes
* space for us first and indicates the fact it did this with this flag.
*/
if(!(execi->userflags & PMEF_AUXVECTORSPACE)) {
char *f = (char *) mysp;
int remain;
vir_bytes extrabytes = sizeof(*a) * PMEF_AUXVECTORS;
/* Create extrabytes more space */
remain = *framelen - (int)(f - frame);
if(*framelen + extrabytes >= ARG_MAX)
return ENOMEM;
*framelen += extrabytes;
AT_SUN_EXECNAME support . vfs: pass execname in aux vectors . ld.elf_so: use this to expand $ORIGIN . this requires the executable to reserve more space at exec() calling time
2012-04-18 16:32:38 +02:00
*newsp -= extrabytes;
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
*extrabase += extrabytes;
memmove(f+extrabytes, f, remain);
memset(f, 0, extrabytes);
}
/* Ok, what mysp points to now we can use for the aux vectors. */
a = (AuxInfo *) mysp;
#define AUXINFO(type, value) \
{ assert((char *) a < (char *) mysp_end); a->a_type = type; a->a_v = value; a++; }
#if 0
AUXINFO(AT_PHENT, elf_header->e_phentsize);
AUXINFO(AT_PHNUM, elf_header->e_phnum);
#endif
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
AUXINFO(AT_BASE, execi->args.load_base);
AUXINFO(AT_ENTRY, execi->args.pc);
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
AUXINFO(AT_PAGESZ, PAGE_SIZE);
AUXINFO(AT_EXECFD, execi->elf_main_fd);
AT_SUN_EXECNAME support . vfs: pass execname in aux vectors . ld.elf_so: use this to expand $ORIGIN . this requires the executable to reserve more space at exec() calling time
2012-04-18 16:32:38 +02:00
/* This is where we add the AT_NULL */
term = a;
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
/* Always terminate with AT_NULL */
AUXINFO(AT_NULL, 0);
AT_SUN_EXECNAME support . vfs: pass execname in aux vectors . ld.elf_so: use this to expand $ORIGIN . this requires the executable to reserve more space at exec() calling time
2012-04-18 16:32:38 +02:00
/* Empty space starts here, if any. */
if((execi->userflags & PMEF_EXECNAMESPACE1)
&& strlen(execi->execname) < PMEF_EXECNAMELEN1) {
char *spacestart;
vir_bytes userp;
/* Make space for the real closing AT_NULL entry. */
AUXINFO(AT_NULL, 0);
/* Empty space starts here; we can put the name here. */
spacestart = (char *) a;
VFS: Use safe string copy functions
2012-07-13 18:08:06 +02:00
strlcpy(spacestart, execi->execname, PATH_MAX);
AT_SUN_EXECNAME support . vfs: pass execname in aux vectors . ld.elf_so: use this to expand $ORIGIN . this requires the executable to reserve more space at exec() calling time
2012-04-18 16:32:38 +02:00
/* What will the address of the string for the user be */
userp = *newsp + (spacestart-frame);
/* Move back to where the AT_NULL is */
a = term;
AUXINFO(AT_SUN_EXECNAME, userp);
AUXINFO(AT_NULL, 0);
}
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
return OK;
}
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
/*===========================================================================*
* is_script *
*===========================================================================*/
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
static int is_script(struct vfs_exec_info *execi)
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
{
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
/* Is Interpreted script? */
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
assert(execi->args.hdr != NULL);
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
return(execi->args.hdr[0] == '#' && execi->args.hdr[1] == '!'
&& execi->args.hdr_len >= 2);
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
}
/*===========================================================================*
* patch_stack *
*===========================================================================*/
retire PUBLIC, PRIVATE and FORWARD
2012-03-25 20:25:53 +02:00
static int patch_stack(vp, stack, stk_bytes, path)
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
struct vnode *vp; /* pointer for open script file */
char stack[ARG_MAX]; /* pointer to stack image within VFS */
VFS: make m_in job local By making m_in job local (i.e., each job has its own copy of m_in instead of refering to the global m_in) we don't have to store and restore m_in on every thread yield. This reduces overhead. Moreover, remove the assumption that m_in is preserved. Do_XXX functions have to copy the system call parameters as soon as possible and only pass those copies to other functions. Furthermore, this patch cleans up some code and uses better types in a lot of places.
2012-04-13 14:50:38 +02:00
size_t *stk_bytes; /* size of initial stack */
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
char path[PATH_MAX]; /* path to script file */
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
{
/* Patch the argument vector to include the path name of the script to be
* interpreted, and all strings on the #! line. Returns the path name of
* the interpreter.
*/
enum { INSERT=FALSE, REPLACE=TRUE };
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
int n, r;
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
off_t pos;
char *sp, *interp = NULL;
New VFS interface
2007-08-07 14:52:47 +02:00
u64_t new_pos;
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
unsigned int cum_io;
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
char buf[_MAX_BLOCK_SIZE];
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
/* Make 'path' the new argv[0]. */
if (!insert_arg(stack, stk_bytes, path, REPLACE)) return(ENOMEM);
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
pos = 0; /* Read from the start of the file */
/* Issue request */
retire 64-bit conversion functions Change-Id: Ib6b81403f877c363a286c654e0524fa1cb781b80
2013-06-17 10:31:12 +02:00
r = req_readwrite(vp->v_fs_e, vp->v_inode_nr, ((u64_t)(pos)), READING,
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
VFS_PROC_NR, buf, _MAX_BLOCK_SIZE, &new_pos, &cum_io);
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
if (r != OK) return(r);
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
n = vp->v_size;
New VFS interface
2007-08-07 14:52:47 +02:00
if (n > _MAX_BLOCK_SIZE)
n = _MAX_BLOCK_SIZE;
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
if (n < 2) return ENOEXEC;
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
sp = &(buf[2]); /* just behind the #! */
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
n -= 2;
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
if (n > PATH_MAX) n = PATH_MAX;
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
/* Use the 'path' variable for temporary storage */
memcpy(path, sp, n);
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
if ((sp = memchr(path, '\n', n)) == NULL) /* must be a proper line */
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
return(ENOEXEC);
/* Move sp backwards through script[], prepending each string to stack. */
for (;;) {
/* skip spaces behind argument. */
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
while (sp > path && (*--sp == ' ' || *sp == '\t')) {}
if (sp == path) break;
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
sp[1] = 0;
/* Move to the start of the argument. */
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
while (sp > path && sp[-1] != ' ' && sp[-1] != '\t') --sp;
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
interp = sp;
vfs - use a static buffer instead of malloc()+free(), solving recently appeared ENOMEM problems during exec().
2010-12-15 15:43:59 +01:00
if (!insert_arg(stack, stk_bytes, sp, INSERT)) {
printf("VFS: patch_stack: insert_arg failed\n");
return(ENOMEM);
}
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
}
vfs,avfs: verify an interpreter was found on #! line . if not, NULL *interp is dereferenced
2011-12-20 16:20:32 +01:00
if(!interp)
return ENOEXEC;
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
/* Round *stk_bytes up to the size of a pointer for alignment contraints. */
*stk_bytes= ((*stk_bytes + PTRSIZE - 1) / PTRSIZE) * PTRSIZE;
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
if (interp != path)
memmove(path, interp, strlen(interp)+1);
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
return(OK);
}
/*===========================================================================*
* insert_arg *
*===========================================================================*/
retire PUBLIC, PRIVATE and FORWARD
2012-03-25 20:25:53 +02:00
static int insert_arg(
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
char stack[ARG_MAX], /* pointer to stack image within PM */
VFS: make m_in job local By making m_in job local (i.e., each job has its own copy of m_in instead of refering to the global m_in) we don't have to store and restore m_in on every thread yield. This reduces overhead. Moreover, remove the assumption that m_in is preserved. Do_XXX functions have to copy the system call parameters as soon as possible and only pass those copies to other functions. Furthermore, this patch cleans up some code and uses better types in a lot of places.
2012-04-13 14:50:38 +02:00
size_t *stk_bytes, /* size of initial stack */
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
char *arg, /* argument to prepend/replace as new argv[0] */
int replace
)
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
{
/* Patch the stack so that arg will become argv[0]. Be careful, the stack may
* be filled with garbage, although it normally looks like this:
* nargs argv[0] ... argv[nargs-1] NULL envp[0] ... NULL
* followed by the strings "pointed" to by the argv[i] and the envp[i]. The
* pointers are really offsets from the start of stack.
* Return true iff the operation succeeded.
*/
VFS: fix new signed/unsigned comparisons
2012-04-02 17:16:44 +02:00
int offset;
vir_bytes a0, a1;
VFS: make m_in job local By making m_in job local (i.e., each job has its own copy of m_in instead of refering to the global m_in) we don't have to store and restore m_in on every thread yield. This reduces overhead. Moreover, remove the assumption that m_in is preserved. Do_XXX functions have to copy the system call parameters as soon as possible and only pass those copies to other functions. Furthermore, this patch cleans up some code and uses better types in a lot of places.
2012-04-13 14:50:38 +02:00
size_t old_bytes = *stk_bytes;
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
/* Prepending arg adds at least one string and a zero byte. */
offset = strlen(arg) + 1;
a0 = (int) ((char **) stack)[1]; /* argv[0] */
if (a0 < 4 * PTRSIZE || a0 >= old_bytes) return(FALSE);
a1 = a0; /* a1 will point to the strings to be moved */
if (replace) {
/* Move a1 to the end of argv[0][] (argv[1] if nargs > 1). */
do {
if (a1 == old_bytes) return(FALSE);
--offset;
} while (stack[a1++] != 0);
} else {
offset += PTRSIZE; /* new argv[0] needs new pointer in argv[] */
a0 += PTRSIZE; /* location of new argv[0][]. */
}
/* stack will grow by offset bytes (or shrink by -offset bytes) */
if ((*stk_bytes += offset) > ARG_MAX) return(FALSE);
/* Reposition the strings by offset bytes */
memmove(stack + a1 + offset, stack + a1, old_bytes - a1);
VFS: Use safe string copy functions
2012-07-13 18:08:06 +02:00
strlcpy(stack + a0, arg, PATH_MAX); /* Put arg in the new space. */
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
if (!replace) {
/* Make space for a new argv[0]. */
memmove(stack + 2 * PTRSIZE, stack + 1 * PTRSIZE, a0 - 2 * PTRSIZE);
((char **) stack)[0]++; /* nargs++; */
}
/* Now patch up argv[] and envp[] by offset. */
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
libexec_patch_ptr(stack, (vir_bytes) offset);
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
((char **) stack)[1] = (char *) a0; /* set argv[0] correctly */
return(TRUE);
}
/*===========================================================================*
* read_seg *
*===========================================================================*/
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
static int read_seg(struct exec_info *execi, off_t off, off_t seg_addr, size_t seg_bytes)
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
{
/*
* The byte count on read is usually smaller than the segment count, because
* a segment is padded out to a click multiple, and the data segment is only
* partially initialized.
*/
Merge of VFS by Balasz Gerofi with Minix trunk.
2006-10-25 15:40:36 +02:00
int r;
New VFS interface
2007-08-07 14:52:47 +02:00
u64_t new_pos;
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
unsigned int cum_io;
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
struct vnode *vp = ((struct vfs_exec_info *) execi->opaque)->vp;
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
/* Make sure that the file is big enough */
VFS: fix new signed/unsigned comparisons
2012-04-02 17:16:44 +02:00
if (off + seg_bytes > LONG_MAX) return(EIO);
if ((unsigned long) vp->v_size < off+seg_bytes) return(EIO);
New VFS interface
2007-08-07 14:52:47 +02:00
retire 64-bit conversion functions Change-Id: Ib6b81403f877c363a286c654e0524fa1cb781b80
2013-06-17 10:31:12 +02:00
if ((r = req_readwrite(vp->v_fs_e, vp->v_inode_nr, ((u64_t)(off)), READING,
No more intel/minix segments. This commit removes all traces of Minix segments (the text/data/stack memory map abstraction in the kernel) and significance of Intel segments (hardware segments like CS, DS that add offsets to all addressing before page table translation). This ultimately simplifies the memory layout and addressing and makes the same layout possible on non-Intel architectures. There are only two types of addresses in the world now: virtual and physical; even the kernel and processes have the same virtual address space. Kernel and user processes can be distinguished at a glance as processes won't use 0xF0000000 and above. No static pre-allocated memory sizes exist any more. Changes to booting: . The pre_init.c leaves the kernel and modules exactly as they were left by the bootloader in physical memory . The kernel starts running using physical addressing, loaded at a fixed location given in its linker script by the bootloader. All code and data in this phase are linked to this fixed low location. . It makes a bootstrap pagetable to map itself to a fixed high location (also in linker script) and jumps to the high address. All code and data then use this high addressing. . All code/data symbols linked at the low addresses is prefixed by an objcopy step with __k_unpaged_*, so that that code cannot reference highly-linked symbols (which aren't valid yet) or vice versa (symbols that aren't valid any more). . The two addressing modes are separated in the linker script by collecting the unpaged_*.o objects and linking them with low addresses, and linking the rest high. Some objects are linked twice, once low and once high. . The bootstrap phase passes a lot of information (e.g. free memory list, physical location of the modules, etc.) using the kinfo struct. . After this bootstrap the low-linked part is freed. . The kernel maps in VM into the bootstrap page table so that VM can begin executing. Its first job is to make page tables for all other boot processes. So VM runs before RS, and RS gets a fully dynamic, VM-managed address space. VM gets its privilege info from RS as usual but that happens after RS starts running. . Both the kernel loading VM and VM organizing boot processes happen using the libexec logic. This removes the last reason for VM to still know much about exec() and vm/exec.c is gone. Further Implementation: . All segments are based at 0 and have a 4 GB limit. . The kernel is mapped in at the top of the virtual address space so as not to constrain the user processes. . Processes do not use segments from the LDT at all; there are no segments in the LDT any more, so no LLDT is needed. . The Minix segments T/D/S are gone and so none of the user-space or in-kernel copy functions use them. The copy functions use a process endpoint of NONE to realize it's a physical address, virtual otherwise. . The umap call only makes sense to translate a virtual address to a physical address now. . Segments-related calls like newmap and alloc_segments are gone. . All segments-related translation in VM is gone (vir2map etc). . Initialization in VM is simpler as no moving around is necessary. . VM and all other boot processes can be linked wherever they wish and will be mapped in at the right location by the kernel and VM respectively. Other changes: . The multiboot code is less special: it does not use mb_print for its diagnostics any more but uses printf() as normal, saving the output into the diagnostics buffer, only printing to the screen using the direct print functions if a panic() occurs. . The multiboot code uses the flexible 'free memory map list' style to receive the list of free memory if available. . The kernel determines the memory layout of the processes to a degree: it tells VM where the kernel starts and ends and where the kernel wants the top of the process to be. VM then uses this entire range, i.e. the stack is right at the top, and mmap()ped bits of memory are placed below that downwards, and the break grows upwards. Other Consequences: . Every process gets its own page table as address spaces can't be separated any more by segments. . As all segments are 0-based, there is no distinction between virtual and linear addresses, nor between userspace and kernel addresses. . Less work is done when context switching, leading to a net performance increase. (8% faster on my machine for 'make servers'.) . The layout and configuration of the GDT makes sysenter and syscall possible.
2012-05-07 16:03:35 +02:00
execi->proc_e, (char*)seg_addr, seg_bytes,
&new_pos, &cum_io)) != OK) {
printf("VFS: read_seg: req_readwrite failed (data)\n");
return(r);
}
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
No more intel/minix segments. This commit removes all traces of Minix segments (the text/data/stack memory map abstraction in the kernel) and significance of Intel segments (hardware segments like CS, DS that add offsets to all addressing before page table translation). This ultimately simplifies the memory layout and addressing and makes the same layout possible on non-Intel architectures. There are only two types of addresses in the world now: virtual and physical; even the kernel and processes have the same virtual address space. Kernel and user processes can be distinguished at a glance as processes won't use 0xF0000000 and above. No static pre-allocated memory sizes exist any more. Changes to booting: . The pre_init.c leaves the kernel and modules exactly as they were left by the bootloader in physical memory . The kernel starts running using physical addressing, loaded at a fixed location given in its linker script by the bootloader. All code and data in this phase are linked to this fixed low location. . It makes a bootstrap pagetable to map itself to a fixed high location (also in linker script) and jumps to the high address. All code and data then use this high addressing. . All code/data symbols linked at the low addresses is prefixed by an objcopy step with __k_unpaged_*, so that that code cannot reference highly-linked symbols (which aren't valid yet) or vice versa (symbols that aren't valid any more). . The two addressing modes are separated in the linker script by collecting the unpaged_*.o objects and linking them with low addresses, and linking the rest high. Some objects are linked twice, once low and once high. . The bootstrap phase passes a lot of information (e.g. free memory list, physical location of the modules, etc.) using the kinfo struct. . After this bootstrap the low-linked part is freed. . The kernel maps in VM into the bootstrap page table so that VM can begin executing. Its first job is to make page tables for all other boot processes. So VM runs before RS, and RS gets a fully dynamic, VM-managed address space. VM gets its privilege info from RS as usual but that happens after RS starts running. . Both the kernel loading VM and VM organizing boot processes happen using the libexec logic. This removes the last reason for VM to still know much about exec() and vm/exec.c is gone. Further Implementation: . All segments are based at 0 and have a 4 GB limit. . The kernel is mapped in at the top of the virtual address space so as not to constrain the user processes. . Processes do not use segments from the LDT at all; there are no segments in the LDT any more, so no LLDT is needed. . The Minix segments T/D/S are gone and so none of the user-space or in-kernel copy functions use them. The copy functions use a process endpoint of NONE to realize it's a physical address, virtual otherwise. . The umap call only makes sense to translate a virtual address to a physical address now. . Segments-related calls like newmap and alloc_segments are gone. . All segments-related translation in VM is gone (vir2map etc). . Initialization in VM is simpler as no moving around is necessary. . VM and all other boot processes can be linked wherever they wish and will be mapped in at the right location by the kernel and VM respectively. Other changes: . The multiboot code is less special: it does not use mb_print for its diagnostics any more but uses printf() as normal, saving the output into the diagnostics buffer, only printing to the screen using the direct print functions if a panic() occurs. . The multiboot code uses the flexible 'free memory map list' style to receive the list of free memory if available. . The kernel determines the memory layout of the processes to a degree: it tells VM where the kernel starts and ends and where the kernel wants the top of the process to be. VM then uses this entire range, i.e. the stack is right at the top, and mmap()ped bits of memory are placed below that downwards, and the break grows upwards. Other Consequences: . Every process gets its own page table as address spaces can't be separated any more by segments. . As all segments are 0-based, there is no distinction between virtual and linear addresses, nor between userspace and kernel addresses. . Less work is done when context switching, leading to a net performance increase. (8% faster on my machine for 'make servers'.) . The layout and configuration of the GDT makes sysenter and syscall possible.
2012-05-07 16:03:35 +02:00
if (r == OK && cum_io != seg_bytes)
printf("VFS: read_seg segment has not been read properly\n");
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
return(r);
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
}
/*===========================================================================*
* clo_exec *
*===========================================================================*/
retire PUBLIC, PRIVATE and FORWARD
2012-03-25 20:25:53 +02:00
static void clo_exec(struct fproc *rfp)
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
{
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
/* Files can be marked with the FD_CLOEXEC bit (in fp->fp_cloexec).
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
*/
. satisfy some gcc warnings (uninitialized/unused variables) . change cloexec mask from long to fd_set to remove 32 fd's per process restriction (from cloexec at least)
2006-06-27 18:47:35 +02:00
int i;
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
- Introduce support for sticky bit. - Revise VFS-FS protocol and update VFS/MFS/ISOFS accordingly. - Clean up MFS by removing old, dead code (backwards compatibility is broken by the new VFS-FS protocol, anyway) and rewrite other parts. Also, make sure all functions have proper banners and prototypes. - VFS should always provide a (syntactically) valid path to the FS; no need for the FS to do sanity checks when leaving/entering mount points. - Fix several bugs in MFS: - Several path lookup bugs in MFS. - A link can be too big for the path buffer. - A mountpoint can become inaccessible when the creation of a new inode fails, because the inode already exists and is a mountpoint. - Introduce support for supplemental groups. - Add test 46 to test supplemental group functionality (and removed obsolete suppl. tests from test 2). - Clean up VFS (not everything is done yet). - ISOFS now opens device read-only. This makes the -r flag in the mount command unnecessary (but will still report to be mounted read-write). - Introduce PipeFS. PipeFS is a new FS that handles all anonymous and named pipes. However, named pipes still reside on the (M)FS, as they are part of the file system on disk. To make this work VFS now has a concept of 'mapped' inodes, which causes read, write, truncate and stat requests to be redirected to the mapped FS, and all other requests to the original FS.
2009-12-20 21:27:14 +01:00
/* Check the file desriptors one by one for presence of FD_CLOEXEC. */
for (i = 0; i < OPEN_MAX; i++)
if ( FD_ISSET(i, &rfp->fp_cloexec_set))
. satisfy some gcc warnings (uninitialized/unused variables) . change cloexec mask from long to fd_set to remove 32 fd's per process restriction (from cloexec at least)
2006-06-27 18:47:35 +02:00
(void) close_fd(rfp, i);
New interface between PM and FS.
2006-05-11 16:57:23 +02:00
}
VFS: replace VFS with AVFS
2012-02-13 16:28:04 +01:00
/*===========================================================================*
* map_header *
*===========================================================================*/
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
static int map_header(struct vfs_exec_info *execi)
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
{
int r;
u64_t new_pos;
unsigned int cum_io;
off_t pos;
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
static char hdr[PAGE_SIZE]; /* Assume that header is not larger than a page */
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
pos = 0; /* Read from the start of the file */
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
/* How much is sensible to read */
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
execi->args.hdr_len = MIN(execi->vp->v_size, sizeof(hdr));
execi->args.hdr = hdr;
recognize and execute dynamically linked executables . generalize libexec slightly to get some more necessary information from ELF files, e.g. the interpreter . execute dynamically linked executables when exec()ed by VFS . switch to netbsd variant of elf32.h exclusively, solves some conflicting headers
2012-04-03 15:52:25 +02:00
r = req_readwrite(execi->vp->v_fs_e, execi->vp->v_inode_nr,
retire 64-bit conversion functions Change-Id: Ib6b81403f877c363a286c654e0524fa1cb781b80
2013-06-17 10:31:12 +02:00
((u64_t)(pos)), READING, VFS_PROC_NR, hdr,
exec() cleanup, generalization, improvement . make exec() callers (i.e. vfs and rs) determine the memory layout by explicitly reserving regions using mmap() calls on behalf of the exec()ing process, i.e. handling all of the exec logic, thereby eliminating all special exec() knowledge from VM. . the new procedure is: clear the exec()ing process first, then call third-party mmap()s to reserve memory, then copy the executable file section contents in, all using callbacks tailored to the caller's way of starting an executable . i.e. no more explicit EXEC_NEWMEM-style calls in PM or VM as with rigid 2-section arguments . this naturally allows generalizing exec() by simply loading all ELF sections . drop/merge of lots of duplicate exec() code into libexec . not copying the code sections to vfs and into the executable again is a measurable performance improvement (about 3.3% faster for 'make' in src/servers/)
2012-05-30 19:34:07 +02:00
execi->args.hdr_len, &new_pos, &cum_io);
vfs - use a static buffer instead of malloc()+free(), solving recently appeared ENOMEM problems during exec().
2010-12-15 15:43:59 +01:00
if (r != OK) {
printf("VFS: exec: map_header: req_readwrite failed\n");
return(r);
}
VFS/RS support for ELF
2010-12-10 10:27:56 +01:00
return(OK);
}
Reference in a new issue Copy permalink