minix/servers/fs/exec.c

/* 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>
#include <minix/callnr.h>
#include <minix/endpoint.h>
#include <minix/com.h>
#include <a.out.h>
#include <signal.h>
#include <string.h>
#include "buf.h"
#include "fproc.h"
#include "inode.h"
#include "param.h"
#include "super.h"

FORWARD _PROTOTYPE( int exec_newmem, (int proc_e, vir_bytes text_bytes,
	vir_bytes data_bytes, vir_bytes bss_bytes, vir_bytes tot_bytes,
	vir_bytes frame_len, int sep_id,
	Dev_t st_dev, ino_t st_ino, time_t st_ctime, char *progname,
	int new_uid, int new_gid,
	vir_bytes *stack_topp, int *load_textp, int *allow_setuidp)	);
FORWARD _PROTOTYPE( int read_header, (struct inode *rip, int *sep_id,
	vir_bytes *text_bytes, vir_bytes *data_bytes,
	vir_bytes *bss_bytes, phys_bytes *tot_bytes, vir_bytes *pc,
	int *hdrlenp)							);
FORWARD _PROTOTYPE( int patch_stack, (struct inode *rip,
	char stack[ARG_MAX], vir_bytes *stk_bytes)			);
FORWARD _PROTOTYPE( int insert_arg, (char stack[ARG_MAX],
	vir_bytes *stk_bytes, char *arg, int replace)			);
FORWARD _PROTOTYPE( void patch_ptr, (char stack[ARG_MAX],
							vir_bytes base)	);
FORWARD _PROTOTYPE( int read_seg, (struct inode *rip, off_t off,
	int proc_e, int seg, phys_bytes seg_bytes)			);
FORWARD _PROTOTYPE( void clo_exec, (struct fproc *rfp)			);

#define ESCRIPT	(-2000)	/* Returned by read_header for a #! script. */
#define PTRSIZE	sizeof(char *) /* Size of pointers in argv[] and envp[]. */

/*===========================================================================*
 *				pm_exec					     *
 *===========================================================================*/
PUBLIC int pm_exec(proc_e, path, path_len, frame, frame_len)
int proc_e;
char *path;
vir_bytes path_len;
char *frame;
vir_bytes frame_len;
{
/* Perform the execve(name, argv, envp) call.  The user library builds a
 * complete stack image, including pointers, args, environ, etc.  The stack
 * is copied to a buffer inside FS, and then to the new core image.
 */
  int r, sep_id, round, proc_s, hdrlen, load_text, allow_setuid;
  vir_bytes text_bytes, data_bytes, bss_bytes, pc;
  phys_bytes tot_bytes;		/* total space for program, including gap */
  vir_bytes stack_top, vsp;
  off_t off;
  uid_t new_uid;
  gid_t new_gid;
  struct fproc *rfp;
  struct inode *rip;
  char *cp;
  char progname[PROC_NAME_LEN];

  static char mbuf[ARG_MAX];	/* buffer for stack and zeroes */

  okendpt(proc_e, &proc_s);
  rfp= fp= &fproc[proc_s];
  who_e= proc_e;
  who_p= proc_s;
  super_user = (fp->fp_effuid == SU_UID ? TRUE : FALSE);   /* su? */

  /* Get the exec file name. */
  r= fetch_name(path, path_len, 0);
  if (r != OK)
  {
	printf("pm_exec: fetch_name failed\n");
	return(r);	/* file name not in user data segment */
  }

  /* Fetch the stack from the user before destroying the old core image. */
  if (frame_len > ARG_MAX)
  {
	printf("pm_exec: bad frame_len\n");
	return(ENOMEM);	/* stack too big */
  }
  r = sys_datacopy(proc_e, (vir_bytes) frame,
  			SELF, (vir_bytes) mbuf, (phys_bytes)frame_len);
  /* can't fetch stack (e.g. bad virtual addr) */
  if (r != OK)
  {
	printf("pm_exec: sys_datacopy failed\n");
	return(r);	
  }

  /* The default is the keep the original user and group IDs */
  new_uid= rfp->fp_effuid;
  new_gid= rfp->fp_effgid;

  for (round= 0; round < 2; round++)
  	/* round = 0 (first attempt), or 1 (interpreted script) */
  {
	/* Save the name of the program */
	(cp= strrchr(user_path, '/')) ? cp++ : (cp= user_path);

	strncpy(progname, cp, PROC_NAME_LEN-1);
	progname[PROC_NAME_LEN-1] = '\0';

#if 0
	printf("pm_exec: eat_path '%s'\n", user_path);
#endif
    	rip= eat_path(user_path);
    	if (rip == NIL_INODE)
	{
		return(err_code);
	}
	if ((rip->i_mode & I_TYPE) != I_REGULAR)
		r = ENOEXEC;
	else
		r = forbidden(rip, X_BIT); /* check if file is executable */
	if (r != OK) {
		put_inode(rip);
		printf("pm_exec: bad executable\n");
		return(r);
	}

	if (round == 0)
	{
		/* Deal with setuid/setgid executables */
		if (rip->i_mode & I_SET_UID_BIT)
			new_uid = rip->i_uid;
		if (rip->i_mode & I_SET_GID_BIT)
			new_gid = rip->i_gid;
	}

	/* Read the file header and extract the segment sizes. */
	r = read_header(rip, &sep_id, &text_bytes, &data_bytes, &bss_bytes, 
		&tot_bytes, &pc, &hdrlen);
	if (r != ESCRIPT || round != 0)
		break;

	/* Get fresh copy of the file name. */
	r= fetch_name(path, path_len, 0);
	if (r != OK)
	{
		printf("pm_exec: 2nd fetch_name failed\n");
		put_inode(rip);
		return(r);	/* strange */
	}
	r= patch_stack(rip, mbuf, &frame_len);
	put_inode(rip);
	if (r != OK)
	{
		printf("pm_exec: patch stack\n");
		return r;
	}
  }

  if (r != OK)
  {
	printf("pm_exec: returning ENOEXEC, r = %d\n", r);
	return ENOEXEC;
  }

  r= exec_newmem(proc_e, text_bytes, data_bytes, bss_bytes, tot_bytes,
	frame_len, sep_id, rip->i_dev, rip->i_num, rip->i_ctime, 
	progname, new_uid, new_gid, &stack_top, &load_text, &allow_setuid);
  if (r != OK)
  {
	printf("pm_exec: exec_newmap failed: %d\n", r);
	put_inode(rip);
	return r;
  }

  /* Patch up stack and copy it from FS to new core image. */
  vsp = stack_top;
  vsp -= frame_len;
  patch_ptr(mbuf, vsp);
  r = sys_datacopy(SELF, (vir_bytes) mbuf,
  		proc_e, (vir_bytes) vsp, (phys_bytes)frame_len);
  if (r != OK) panic(__FILE__,"pm_exec stack copy err on", proc_e);

  off = hdrlen;

  /* Read in text and data segments. */
  if (load_text) {
	r= read_seg(rip, off, proc_e, T, text_bytes);
  }
  off += text_bytes;
  if (r == OK)
  	r= read_seg(rip, off, proc_e, D, data_bytes);

  put_inode(rip);

  if (r != OK) return r;

  clo_exec(rfp);

  if (allow_setuid)
  {
	rfp->fp_effuid= new_uid;
	rfp->fp_effgid= new_gid;
  }

  /* This child has now exec()ced. */
  rfp->fp_execced = 1;

  /* Check if this is a driver that can now be useful. */
  dmap_endpt_up(rfp->fp_endpoint);

  return OK;
}


/*===========================================================================*
 *				exec_newmem				     *
 *===========================================================================*/
PRIVATE int exec_newmem(proc_e, text_bytes, data_bytes, bss_bytes, tot_bytes,
	frame_len, sep_id, st_dev, st_ino, st_ctime, progname,
	new_uid, new_gid, stack_topp, load_textp, allow_setuidp)
int proc_e;
vir_bytes text_bytes;
vir_bytes data_bytes;
vir_bytes bss_bytes;
vir_bytes tot_bytes;
vir_bytes frame_len;
int sep_id;
dev_t st_dev;
ino_t st_ino;
time_t st_ctime;
int new_uid;
int new_gid;
char *progname;
vir_bytes *stack_topp;
int *load_textp;
int *allow_setuidp;
{
	int r;
	struct exec_newmem e;
	message m;

	e.text_bytes= text_bytes;
	e.data_bytes= data_bytes;
	e.bss_bytes= bss_bytes;
	e.tot_bytes= tot_bytes;
	e.args_bytes= frame_len;
	e.sep_id= sep_id;
	e.st_dev= st_dev;
	e.st_ino= st_ino;
	e.st_ctime= st_ctime;
	e.new_uid= new_uid;
	e.new_gid= new_gid;
	strncpy(e.progname, progname, sizeof(e.progname)-1);
	e.progname[sizeof(e.progname)-1]= '\0';

	m.m_type= EXEC_NEWMEM;
	m.EXC_NM_PROC= proc_e;
	m.EXC_NM_PTR= (char *)&e;
	r= sendrec(PM_PROC_NR, &m);
	if (r != OK)
		return r;
#if 0
	printf("exec_newmem: r = %d, m_type = %d\n", r, m.m_type);
#endif
	*stack_topp= m.m1_i1;
	*load_textp= !!(m.m1_i2 & EXC_NM_RF_LOAD_TEXT);
	*allow_setuidp= !!(m.m1_i2 & EXC_NM_RF_ALLOW_SETUID);
#if 0
	printf("exec_newmem: stack_top = 0x%x\n", *stack_topp);
	printf("exec_newmem: load_text = %d\n", *load_textp);
#endif
	return m.m_type;
}


/*===========================================================================*
 *				read_header				     *
 *===========================================================================*/
PRIVATE int read_header(rip, sep_id, text_bytes, data_bytes, bss_bytes, 
						tot_bytes, pc, hdrlenp)
struct inode *rip;		/* inode for reading exec file */
int *sep_id;			/* true iff sep I&D */
vir_bytes *text_bytes;		/* place to return text size */
vir_bytes *data_bytes;		/* place to return initialized data size */
vir_bytes *bss_bytes;		/* place to return bss size */
phys_bytes *tot_bytes;		/* place to return total size */
vir_bytes *pc;			/* program entry point (initial PC) */
int *hdrlenp;
{
/* Read the header and extract the text, data, bss and total sizes from it. */
  off_t pos;
  block_t b;
  struct buf *bp;
  struct exec hdr;		/* a.out header is read in here */

  /* Read the header and check the magic number.  The standard MINIX header 
   * is defined in <a.out.h>.  It consists of 8 chars followed by 6 longs.
   * Then come 4 more longs that are not used here.
   *	Byte 0: magic number 0x01
   *	Byte 1: magic number 0x03
   *	Byte 2: normal = 0x10 (not checked, 0 is OK), separate I/D = 0x20
   *	Byte 3: CPU type, Intel 16 bit = 0x04, Intel 32 bit = 0x10, 
   *            Motorola = 0x0B, Sun SPARC = 0x17
   *	Byte 4: Header length = 0x20
   *	Bytes 5-7 are not used.
   *
   *	Now come the 6 longs
   *	Bytes  8-11: size of text segments in bytes
   *	Bytes 12-15: size of initialized data segment in bytes
   *	Bytes 16-19: size of bss in bytes
   *	Bytes 20-23: program entry point
   *	Bytes 24-27: total memory allocated to program (text, data + stack)
   *	Bytes 28-31: size of symbol table in bytes
   * The longs are represented in a machine dependent order,
   * little-endian on the 8088, big-endian on the 68000.
   * The header is followed directly by the text and data segments, and the 
   * symbol table (if any). The sizes are given in the header. Only the 
   * text and data segments are copied into memory by exec. The header is 
   * used here only. The symbol table is for the benefit of a debugger and 
   * is ignored here.
   */

  pos= 0;	/* Read from the start of the file */
  b = read_map(rip, pos);	/* get block number */

  if (b == 0)	/* Hole */
	return ENOEXEC;

  bp = get_block(rip->i_dev, b, NORMAL);	/* get block */

  /* Interpreted script? */
  if (bp->b_data[0] == '#' && bp->b_data[1] == '!' && rip->i_size >= 2)
  {
	put_block(bp, FULL_DATA_BLOCK);
	return ESCRIPT;
  }

  memcpy(&hdr, bp->b_data, sizeof(hdr));
  put_block(bp, FULL_DATA_BLOCK);

  if (rip->i_size < A_MINHDR) return(ENOEXEC);

  /* Check magic number, cpu type, and flags. */
  if (BADMAG(hdr)) return(ENOEXEC);
#if (CHIP == INTEL && _WORD_SIZE == 2)
  if (hdr.a_cpu != A_I8086) return(ENOEXEC);
#endif
#if (CHIP == INTEL && _WORD_SIZE == 4)
  if (hdr.a_cpu != A_I80386) return(ENOEXEC);
#endif
  if ((hdr.a_flags & ~(A_NSYM | A_EXEC | A_SEP)) != 0) return(ENOEXEC);

  *sep_id = !!(hdr.a_flags & A_SEP);	    /* separate I & D or not */

  /* Get text and data sizes. */
  *text_bytes = (vir_bytes) hdr.a_text;	/* text size in bytes */
  *data_bytes = (vir_bytes) hdr.a_data;	/* data size in bytes */
  *bss_bytes  = (vir_bytes) hdr.a_bss;	/* bss size in bytes */
  *tot_bytes  = hdr.a_total;		/* total bytes to allocate for prog */
  if (*tot_bytes == 0) return(ENOEXEC);

  if (!*sep_id) {
	/* If I & D space is not separated, it is all considered data. Text=0*/
	*data_bytes += *text_bytes;
	*text_bytes = 0;
  }
  *pc = hdr.a_entry;	/* initial address to start execution */
  *hdrlenp = hdr.a_hdrlen & BYTE;		/* header length */

  return(OK);
}

/*===========================================================================*
 *				patch_stack				     *
 *===========================================================================*/
PRIVATE int patch_stack(rip, stack, stk_bytes)
struct inode *rip;		/* pointer for open script file */
char stack[ARG_MAX];		/* pointer to stack image within FS */
vir_bytes *stk_bytes;		/* size of initial stack */
{
/* 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 };
  int n;
  off_t pos;
  block_t b;
  struct buf *bp;
  char *sp, *interp = NULL;

  /* Make user_path the new argv[0]. */
  if (!insert_arg(stack, stk_bytes, user_path, REPLACE)) return(ENOMEM);

  pos= 0;	/* Read from the start of the file */
  b = read_map(rip, pos);	/* get block number */
  if (b == 0)	/* Hole */
	return ENOEXEC;

  bp = get_block(rip->i_dev, b, NORMAL);	/* get block */
  n= rip->i_size;
  if (n > rip->i_sp->s_block_size)
	n= rip->i_sp->s_block_size;
  if (n < 2)
  {
	put_block(bp, FULL_DATA_BLOCK);
	return ENOEXEC;
  }
  sp= bp->b_data+2;				/* just behind the #! */
  n -= 2;
  if (n > PATH_MAX)
	n= PATH_MAX;

  /* Use the user_path variable for temporary storage */
  memcpy(user_path, sp, n);
  put_block(bp, FULL_DATA_BLOCK);

  if ((sp= memchr(user_path, '\n', n)) == NULL)	/* must be a proper line */
	return(ENOEXEC);

  /* Move sp backwards through script[], prepending each string to stack. */
  for (;;) {
	/* skip spaces behind argument. */
	while (sp > user_path && (*--sp == ' ' || *sp == '\t')) {}
	if (sp == user_path) break;

	sp[1] = 0;
	/* Move to the start of the argument. */
	while (sp > user_path && sp[-1] != ' ' && sp[-1] != '\t') --sp;

	interp = sp;
	if (!insert_arg(stack, stk_bytes, sp, INSERT)) return(ENOMEM);
  }

  /* Round *stk_bytes up to the size of a pointer for alignment contraints. */
  *stk_bytes= ((*stk_bytes + PTRSIZE - 1) / PTRSIZE) * PTRSIZE;

  if (interp != user_path)
	memmove(user_path, interp, strlen(interp)+1);
  return(OK);
}

/*===========================================================================*
 *				insert_arg				     *
 *===========================================================================*/
PRIVATE int insert_arg(stack, stk_bytes, arg, replace)
char stack[ARG_MAX];		/* pointer to stack image within PM */
vir_bytes *stk_bytes;		/* size of initial stack */
char *arg;			/* argument to prepend/replace as new argv[0] */
int replace;
{
/* 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.
 */
  int offset, a0, a1, old_bytes = *stk_bytes;

  /* 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);

  strcpy(stack + a0, arg);	/* Put arg in the new space. */

  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. */
  patch_ptr(stack, (vir_bytes) offset);
  ((char **) stack)[1] = (char *) a0;	/* set argv[0] correctly */
  return(TRUE);
}

/*===========================================================================*
 *				patch_ptr				     *
 *===========================================================================*/
PRIVATE void patch_ptr(stack, base)
char stack[ARG_MAX];		/* pointer to stack image within PM */
vir_bytes base;			/* virtual address of stack base inside user */
{
/* When doing an exec(name, argv, envp) call, the user builds up a stack
 * image with arg and env pointers relative to the start of the stack.  Now
 * these pointers must be relocated, since the stack is not positioned at
 * address 0 in the user's address space.
 */

  char **ap, flag;
  vir_bytes v;

  flag = 0;			/* counts number of 0-pointers seen */
  ap = (char **) stack;		/* points initially to 'nargs' */
  ap++;				/* now points to argv[0] */
  while (flag < 2) {
	if (ap >= (char **) &stack[ARG_MAX]) return;	/* too bad */
	if (*ap != NULL) {
		v = (vir_bytes) *ap;	/* v is relative pointer */
		v += base;		/* relocate it */
		*ap = (char *) v;	/* put it back */
	} else {
		flag++;
	}
	ap++;
  }
}

/*===========================================================================*
 *				read_seg				     *
 *===========================================================================*/
PRIVATE int read_seg(rip, off, proc_e, seg, seg_bytes)
struct inode *rip; 		/* inode descriptor to read from */
off_t off;			/* offset in file */
int proc_e;			/* process number (endpoint) */
int seg;			/* T, D, or S */
phys_bytes seg_bytes;		/* how much is to be transferred? */
{
/*
 * 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.
 */

  int r, block_size;
  off_t n, o, b_off, seg_off;
  block_t b;
  struct buf *bp;

  /* Make sure that the file is big enough */
  if (rip->i_size < off+seg_bytes)
	return EIO;

  block_size= rip->i_sp->s_block_size;
  seg_off= 0;
  for (o= off - (off % block_size); o < off+seg_bytes; o += block_size)
  {
	b= read_map(rip, o);
	if (b == NO_BLOCK)
		return EIO;	/* Executables don't have holes */

	bp = get_block(rip->i_dev, b, NORMAL);	/* get block */
	if (o < off)
		b_off= off-o;
	else
		b_off= 0;
	n= block_size-b_off;
	if (o+b_off+n > off+seg_bytes)
		n= off+seg_bytes-(o+b_off);
	r= sys_vircopy(SELF, D, (vir_bytes)bp->b_data+b_off,
		proc_e, seg, seg_off, n);
	put_block(bp, FULL_DATA_BLOCK);

	if (r != OK)
		return r;

	seg_off += n;
  }

  return OK;
}


/*===========================================================================*
 *				clo_exec				     *
 *===========================================================================*/
PRIVATE void clo_exec(rfp)
struct fproc *rfp;
{
/* Files can be marked with the FD_CLOEXEC bit (in fp->fp_cloexec).  
 */
  int i, proc;
  long bitmap;

  /* The array of FD_CLOEXEC bits is in the fp_cloexec bit map. */
  bitmap = rfp->fp_cloexec;
  if (bitmap) {
    /* Check the file desriptors one by one for presence of FD_CLOEXEC. */
    for (i = 0; i < OPEN_MAX; i++) {
	  if ( (bitmap >> i) & 01) (void) close_fd(rfp, i);
    }
  }
}