path: root/procfs/process.c

/* Hurd /proc filesystem, implementation of process directories.
   Copyright (C) 2010,14 Free Software Foundation, Inc.

   This file is part of the GNU Hurd.

   The GNU Hurd is free software; you can redistribute it and/or
   modify it under the terms of the GNU General Public License as
   published by the Free Software Foundation; either version 2, or (at
   your option) any later version.

   The GNU Hurd is distributed in the hope that it will be useful, but
   WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
   General Public License for more details.

   You should have received a copy of the GNU General Public License
   along with this program; if not, write to the Free Software
   Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */

#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <hurd/process.h>
#include <hurd/resource.h>
#include <mach/vm_param.h>
#include <ps.h>
#include "procfs.h"
#include "procfs_dir.h"
#include "process.h"
#include "main.h"

/* This module implements the process directories and the files they
   contain.  A libps proc_stat structure is created for each process
   node, and is used by the individual file content generators as a
   source of information.  Each possible file (cmdline, environ, ...) is
   described in a process_file_desc structure, which specifies which bits
   of information (ie. libps flags) it needs, and what function should
   be used to generate the file's contents.

   The content generators are defined first, followed by glue logic and
   entry table.  */


/* Helper functions */

static char state_char (struct proc_stat *ps)
{
  int i;

  for (i = 0; (1 << i) & (PSTAT_STATE_P_STATES | PSTAT_STATE_T_STATES); i++)
    if (proc_stat_state (ps) & (1 << i))
      return proc_stat_state_tags[i];

  return '?';
}

static const char *state_string (struct proc_stat *ps)
{
  static const char *const state_strings[] = {
    "T (stopped)",
    "Z (zombie)",
    "R (running)",
    "H (halted)",
    "D (disk sleep)",
    "S (sleeping)",
    "I (idle)",
    NULL
  };
  int i;

  for (i = 0; state_strings[i]; i++)
    if (proc_stat_state (ps) & (1 << i))
      return state_strings[i];

  return "? (unknown)";
}

static long long int timeval_jiffies (time_value_t tv)
{
  double secs = tv.seconds * 1000000. + tv.microseconds;
  return secs * opt_clk_tck / 1000000.;
}

static const char *args_filename (const char *name)
{
  char *sp = strrchr (name, '/');
  return sp != NULL && *(sp + 1) != '\0' ? sp + 1 : name;
}

static int args_filename_length (const char *name)
{
  return strchrnul (name, ' ') - name;
}

/* Actual content generators */

static ssize_t
process_file_gc_cmdline (struct proc_stat *ps, char **contents)
{
  *contents = proc_stat_args(ps);
  return proc_stat_args_len(ps);
}

static ssize_t
process_file_gc_environ (struct proc_stat *ps, char **contents)
{
  *contents = proc_stat_env(ps);
  return proc_stat_env_len(ps);
}

static ssize_t
process_file_gc_maps (struct proc_stat *ps, char **contents)
{
  error_t err;
  FILE *s;
  size_t contents_len;
  vm_offset_t addr = 0;
  vm_size_t size;
  vm_prot_t prot, max_prot;
  mach_port_t obj;
  vm_offset_t offs;
  vm_inherit_t inh;
  int shared;

  /* Unfortunately we cannot resolve memory objects to their backing
     file (yet), so we use the port name as identifier.  To avoid the
     same name from being used again and again, we defer the
     deallocation until the end of the function.  We use a simple
     linked list for this purpose.  */
  struct mem_obj
    {
      mach_port_t port;
      struct mem_obj *next;
    };
  struct mem_obj *objects = NULL;

  s = open_memstream (contents, &contents_len);
  if (s == NULL)
    {
      *contents = NULL;
      return 0;
    }

  while (1)
    {
      err =
	vm_region (ps->task, &addr, &size, &prot, &max_prot, &inh,
		   &shared, &obj, &offs);
      if (err)
	break;

      fprintf (s, "%0*lx-%0*lx %c%c%c%c %0*lx %s %d ",
	       /* Address range.  */
	       2*sizeof s, addr,
	       2*sizeof s, addr + size,
	       /* Permissions.	*/
	       prot & VM_PROT_READ? 'r': '-',
	       prot & VM_PROT_WRITE? 'w': '-',
	       prot & VM_PROT_EXECUTE? 'x': '-',
	       shared? 's': 'p',
	       /* Offset.  */
	       2*sizeof s, offs,
	       /* Device.  */
	       "00:00",
	       /* Inode.  */
	       0);

      /* Pathname.  */
      if (MACH_PORT_VALID (obj))
	{
	  struct mem_obj *o = malloc (sizeof *o);
	  if (o)
	    {
	      o->port = obj;
	      o->next = objects;
	      objects = o;
	    }
	  else
	    mach_port_deallocate (mach_task_self (), obj);

	  fprintf (s, "[mem_obj=%lu]\n", obj);
	}
      else
	fprintf (s, "\n");

      addr += size;
    }

  while (objects)
    {
      struct mem_obj *o = objects;
      mach_port_deallocate (mach_task_self (), o->port);
      objects = o->next;
      free (o);
    }

  /* This is a bit awkward, fortunately vm_region should not fail.  */
  if (err != KERN_NO_SPACE)
    fprintf (s, "%s\n", strerror (err));

  fclose (s);
  return contents_len;
}

static ssize_t
process_file_gc_stat (struct proc_stat *ps, char **contents)
{
  struct procinfo *pi = proc_stat_proc_info (ps);
  task_basic_info_t tbi = proc_stat_task_basic_info (ps);
  thread_basic_info_t thbi = proc_stat_thread_basic_info (ps);
  const char *fn = args_filename (proc_stat_args (ps));

  vm_address_t start_code = 1; /* 0 would make killall5.c consider it
				  a kernel process, thus use 1 as
				  default.  */
  vm_address_t end_code = 1;
  process_t p;
  error_t err = proc_pid2proc (ps->context->server, ps->pid, &p);
  if (! err)
    {
      boolean_t essential = 0;
      proc_is_important (p, &essential);
      if (essential)
	start_code = end_code = 0; /* To make killall5.c consider it a
				      kernel process that is to be
				      left alone.  */
      else
	proc_get_code (p, &start_code, &end_code);

      mach_port_deallocate (mach_task_self (), p);
    }

  /* See proc(5) for more information about the contents of each field for the
     Linux procfs.  */
  return asprintf (contents,
      "%d (%.*s) %c "		/* pid, command, state */
      "%d %d %d "		/* ppid, pgid, session */
      "%d %d "			/* controlling tty stuff */
      "%u "			/* flags, as defined by <linux/sched.h> */
      "%lu %lu %lu %lu "	/* page fault counts */
      "%lu %lu %ld %ld "	/* user/sys times, in sysconf(_SC_CLK_TCK) */
      "%d %d "			/* scheduler params (priority, nice) */
      "%d %ld "			/* number of threads, [obsolete] */
      "%llu "			/* start time since boot (jiffies) */
      "%lu %ld %lu "		/* virtual size (bytes), rss (pages), rss lim */
      "%lu %lu %lu %lu %lu "	/* some vm addresses (code, stack, sp, pc) */
      "%lu %lu %lu %lu "	/* pending, blocked, ignored and caught sigs */
      "%lu "			/* wait channel */
      "%lu %lu "		/* swap usage (not maintained in Linux) */
      "%d "			/* exit signal, to be sent to the parent */
      "%d "			/* last processor used */
      "%u %u "			/* RT priority and policy */
      "%llu "			/* aggregated block I/O delay */
      "\n",
      proc_stat_pid (ps), args_filename_length (fn), fn, state_char (ps),
      pi->ppid, pi->pgrp, pi->session,
      0, 0,		/* no such thing as a major:minor for ctty */
      0,		/* no such thing as CLONE_* flags on Hurd */
      0L, 0L, 0L, 0L,	/* TASK_EVENTS_INFO is unavailable on GNU Mach */
      (long unsigned) timeval_jiffies (thbi->user_time),
      (long unsigned) timeval_jiffies (thbi->system_time),
      0L, 0L,		/* cumulative time for children */
      MACH_PRIORITY_TO_NICE(thbi->base_priority) + 20,
      MACH_PRIORITY_TO_NICE(thbi->base_priority),
      pi->nthreads, 0L,
      timeval_jiffies (thbi->creation_time), /* FIXME: ... since boot */
      (long unsigned) tbi->virtual_size,
      (long unsigned) tbi->resident_size / PAGE_SIZE, 0L,
      start_code,
      end_code,
      0L, 0L, 0L,
      0L, 0L, 0L, 0L,
      (long unsigned) proc_stat_thread_rpc (ps), /* close enough */
      0L, 0L,
      0,
      0,
      0, 0,
      0LL);
}

static ssize_t
process_file_gc_statm (struct proc_stat *ps, char **contents)
{
  task_basic_info_t tbi = proc_stat_task_basic_info (ps);

  return asprintf (contents,
      "%lu %lu 0 0 0 0 0\n",
      tbi->virtual_size  / sysconf(_SC_PAGE_SIZE),
      tbi->resident_size / sysconf(_SC_PAGE_SIZE));
}

static ssize_t
process_file_gc_status (struct proc_stat *ps, char **contents)
{
  task_basic_info_t tbi = proc_stat_task_basic_info (ps);
  const char *fn = args_filename (proc_stat_args (ps));

  return asprintf (contents,
      "Name:\t%.*s\n"
      "State:\t%s\n"
      "Tgid:\t%u\n"
      "Pid:\t%u\n"
      "PPid:\t%u\n"
      "Uid:\t%u\t%u\t%u\t%u\n"
      "VmSize:\t%8u kB\n"
      "VmPeak:\t%8u kB\n"
      "VmRSS:\t%8u kB\n"
      "VmHWM:\t%8u kB\n" /* ie. resident peak */
      "Threads:\t%u\n",
      args_filename_length (fn), fn,
      state_string (ps),
      proc_stat_pid (ps), /* XXX will need more work for threads */
      proc_stat_pid (ps),
      proc_stat_proc_info (ps)->ppid,
      proc_stat_owner_uid (ps),
      proc_stat_owner_uid (ps),
      proc_stat_owner_uid (ps),
      proc_stat_owner_uid (ps),
      tbi->virtual_size / 1024,
      tbi->virtual_size / 1024,
      tbi->resident_size / 1024,
      tbi->resident_size / 1024,
      proc_stat_num_threads (ps));
}


/* Implementation of the file nodes. */

/* Describes a file in the process directories.  This structure is
   filled in as an "entry hook" in our procfs_dir entry table and is
   passed to the process_file_make_node function defined below.  */
struct process_file_desc
{
  /* The proc_stat information required to get the contents of this file.  */
  ps_flags_t needs;

  /* Content generator to use for this file.  Once we have acquired the
     necessary information, there can be only memory allocation errors,
     hence this simplified signature.  */
  ssize_t (*get_contents) (struct proc_stat *ps, char **contents);

  /* The cmdline and environ contents don't need any cleaning since they
     point directly into the proc_stat structure.  */
  int no_cleanup;

  /* If specified, the file mode to be set with procfs_node_chmod().  */
  mode_t mode;
};

struct process_file_node
{
  const struct process_file_desc *desc;
  struct proc_stat *ps;
};

/* FIXME: lock the parent! */
static error_t
process_file_get_contents (void *hook, char **contents, ssize_t *contents_len)
{
  struct process_file_node *file = hook;
  error_t err;

  /* Fetch the required information.  */
  err = proc_stat_set_flags (file->ps, file->desc->needs);
  if (err)
    return EIO;
  if ((proc_stat_flags (file->ps) & file->desc->needs) != file->desc->needs)
    return EIO;

  /* Call the actual content generator (see the definitions below).  */
  *contents_len = file->desc->get_contents (file->ps, contents);
  return 0;
}

static void
process_file_cleanup_contents (void *hook, char *contents, ssize_t len)
{
  struct process_file_node *file = hook;

  if (! file->desc->no_cleanup)
    free (contents);
}

static struct node *
process_file_make_node (void *dir_hook, const void *entry_hook)
{
  static const struct procfs_node_ops ops = {
    .get_contents = process_file_get_contents,
    .cleanup_contents = process_file_cleanup_contents,
    .cleanup = free,
  };
  struct process_file_node *f;
  struct node *np;

  f = malloc (sizeof *f);
  if (! f)
    return NULL;

  f->desc = entry_hook;
  f->ps = dir_hook;

  np = procfs_make_node (&ops, f);
  if (! np)
    return NULL;

  procfs_node_chown (np, proc_stat_owner_uid (f->ps));
  if (f->desc->mode)
    procfs_node_chmod (np, f->desc->mode);

  return np;
}

/* Stat needs its own constructor in order to set its mode according to
   the --stat-mode command-line option.  */
static struct node *
process_stat_make_node (void *dir_hook, const void *entry_hook)
{
  struct node *np = process_file_make_node (dir_hook, entry_hook);
  if (np) procfs_node_chmod (np, opt_stat_mode);
  return np;
}


/* Implementation of the process directory per se.  */

static struct procfs_dir_entry entries[] = {
  {
    .name = "cmdline",
    .hook = & (struct process_file_desc) {
      .get_contents = process_file_gc_cmdline,
      .needs = PSTAT_ARGS,
      .no_cleanup = 1,
    },
  },
  {
    .name = "environ",
    .hook = & (struct process_file_desc) {
      .get_contents = process_file_gc_environ,
      .needs = PSTAT_ENV,
      .no_cleanup = 1,
      .mode = 0400,
    },
  },
  {
    .name = "maps",
    .hook = & (struct process_file_desc) {
      .get_contents = process_file_gc_maps,
      .needs = PSTAT_TASK,
      .mode = 0400,
    },
  },
  {
    .name = "stat",
    .hook = & (struct process_file_desc) {
      .get_contents = process_file_gc_stat,
      .needs = PSTAT_PID | PSTAT_ARGS | PSTAT_STATE | PSTAT_PROC_INFO
	| PSTAT_TASK | PSTAT_TASK_BASIC | PSTAT_THREAD_BASIC
	| PSTAT_THREAD_WAIT,
    },
    .ops = {
      .make_node = process_stat_make_node,
    }
  },
  {
    .name = "statm",
    .hook = & (struct process_file_desc) {
      .get_contents = process_file_gc_statm,
      .needs = PSTAT_TASK_BASIC,
    },
  },
  {
    .name = "status",
    .hook = & (struct process_file_desc) {
      .get_contents = process_file_gc_status,
      .needs = PSTAT_PID | PSTAT_ARGS | PSTAT_STATE | PSTAT_PROC_INFO
        | PSTAT_TASK_BASIC | PSTAT_OWNER_UID | PSTAT_NUM_THREADS,
    },
  },
  {}
};

error_t
process_lookup_pid (struct ps_context *pc, pid_t pid, struct node **np)
{
  static const struct procfs_dir_ops dir_ops = {
    .entries = entries,
    .cleanup = (void (*)(void *)) _proc_stat_free,
    .entry_ops = {
      .make_node = process_file_make_node,
    },
  };
  struct proc_stat *ps;
  int owner;
  error_t err;

  err = _proc_stat_create (pid, pc, &ps);
  if (err == ESRCH)
    return ENOENT;
  if (err)
    return EIO;

  err = proc_stat_set_flags (ps, PSTAT_OWNER_UID);
  if (err || ! (proc_stat_flags (ps) & PSTAT_OWNER_UID))
    {
      _proc_stat_free (ps);
      return EIO;
    }

  *np = procfs_dir_make_node (&dir_ops, ps);
  if (! *np)
    return ENOMEM;

  owner = proc_stat_owner_uid (ps);
  procfs_node_chown (*np, owner >= 0 ? owner : opt_anon_owner);
  return 0;
}