Sender: rich AT phekda DOT freeserve DOT co DOT uk
Message-ID: <3BE2B26D.CA65166B@phekda.freeserve.co.uk>
Date: Fri, 02 Nov 2001 14:49:17 +0000
From: Richard Dawe <rich AT phekda DOT freeserve DOT co DOT uk>
X-Mailer: Mozilla 4.77 [en] (X11; U; Linux 2.2.19 i586)
X-Accept-Language: de,fr
MIME-Version: 1.0
To: DJGPP workers <djgpp-workers AT delorie DOT com>
Subject: Patch to add st_blocks to struct stat
References: <Pine DOT SUN DOT 3 DOT 91 DOT 1000713092214 DOT 12595A-100000 AT is>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com

Hello.

A long time ago in a galaxy very, very close to here, Eli Zaretskii wrote:
> The logical step would be to add code to `stat' that computes the
> st_blocks field (it is not in `struct stat', so you will need to
> change the definition of `struct stat' as well).  This could be done
> by computing the cluster size for each drive letter the first time you
> see that drive, and caching that value for future use.  Then you can
> round up the file's size using that cluster size.
> 
> Once we support st_blocks, the changes to `du' to use it would be
> trivial.  Ideally, the `configure' script should detect that and cause
> `du' to use it automagically.

Please find below a patch that adds and calculates st_blocks in struct
stat. Please note that the patch to fd_props.h may report some fuzz, when
applied. I hacked out a small diff, which I had added when I started
trying to add LFS support to DJGPP.

Here's how the patch works:

xstat.c now has a cache for storing the block size on various drive
letters. The first time a file on, say, C: is stat'd, statvfs is called on
the filename. The resulting block size is stored and then returned for all
files on C: in the future. Note that _fixpath is called in the caching
mechanism, so it should cope with joined drives, I think.

I added a function to fd_props.h to return the filename, given a handle.
This is needed in fstat, so that we know which drive the file is on.

One problem I can think of is that there is no aging out in the cache. So
if we map and unmap a network drive during the execution of a program
which has stat'd the original mapping, the program will not notice. Is
this really a problem?

Now to the original motivation: fileutils does detect that we have
st_blocks and then it uses it. Unfortunately du 4.0 rounds file sizes
down, so if you have a big block size (like I seem to on my HD) and you
ask for a size shown in <something>bytes, then the file sizes are bogus
(e.g. 0 for files < 16K on my HD). du 4.1 rounds up file sizes, so it
should be OK.

Same st_blocks values:

My 24GB DOS partition: 16 Kbytes
Mapped network drive (Samba share from Linux box): 32 Kbytes
1.44MB floppy: 512 bytes

Oh yeah, I also fixed the *stat test programs a bit - include needed
headers, return values from main - and made them display block
information.

OK to commit?

Thanks, bye, Rich =]

-- 
Richard Dawe

http://www.phekda.freeserve.co.uk/richdawe/
Index: include/sys/types.h
===================================================================
RCS file: /cvs/djgpp/djgpp/include/sys/types.h,v
retrieving revision 1.6
diff -p -c -3 -r1.6 types.h
*** include/sys/types.h 2000/12/05 14:05:53     1.6
--- include/sys/types.h 2001/11/02 14:36:20
*************** extern "C" {
*** 13,19 ****
  #ifndef __STRICT_ANSI__
  
  #include <sys/djtypes.h>
!   
  typedef int           dev_t;
  typedef int           ino_t;
  typedef int           mode_t;
--- 13,21 ----
  #ifndef __STRICT_ANSI__
  
  #include <sys/djtypes.h>
! 
! typedef int             blkcnt_t;
! typedef int             blksize_t;
  typedef int           dev_t;
  typedef int           ino_t;
  typedef int           mode_t;
Index: include/sys/stat.h
===================================================================
RCS file: /cvs/djgpp/djgpp/include/sys/stat.h,v
retrieving revision 1.4
diff -p -c -3 -r1.4 stat.h
*** include/sys/stat.h  2000/12/05 14:05:53     1.4
--- include/sys/stat.h  2001/11/02 14:36:24
*************** struct stat {
*** 49,57 ****
    time_t      st_mtime;
    nlink_t     st_nlink;
    off_t               st_size;
!   off_t               st_blksize;
    uid_t               st_uid;
    dev_t               st_rdev; /* unused */
  };
  
  int   chmod(const char *_path, mode_t _mode);
--- 49,58 ----
    time_t      st_mtime;
    nlink_t     st_nlink;
    off_t               st_size;
!   blksize_t   st_blksize;
    uid_t               st_uid;
    dev_t               st_rdev; /* unused */
+   blkcnt_t      st_blocks;
  };
  
  int   chmod(const char *_path, mode_t _mode);
Index: include/libc/fd_props.h
===================================================================
RCS file: /cvs/djgpp/djgpp/include/libc/fd_props.h,v
retrieving revision 1.6
diff -p -c -3 -r1.6 fd_props.h
*** include/libc/fd_props.h     2001/06/06 21:09:50     1.6
--- include/libc/fd_props.h     2001/11/02 14:36:27
*************** extern "C" {
*** 8,13 ****
--- 8,15 ----
  
  #ifndef __dj_ENFORCE_ANSI_FREESTANDING
  
+ #include <string.h>
+ 
  #ifndef __STRICT_ANSI__
  
  #ifndef _POSIX_SOURCE
*************** static __inline__ void __clear_fd_flags(
*** 63,68 ****
--- 68,78 ----
  static __inline__ unsigned long __get_fd_flags(int _fd)
  {
    return __has_fd_properties(_fd) ? __fd_properties[_fd]->flags : 0;
+ }
+ 
+ static __inline__ const char * __get_fd_name(int _fd)
+ {
+   return __has_fd_properties(_fd) ? __fd_properties[_fd]->filename :
NULL;
  }
  
  #endif /* !_POSIX_SOURCE */
Index: src/libc/posix/sys/stat/fstat.c
===================================================================
RCS file: /cvs/djgpp/djgpp/src/libc/posix/sys/stat/fstat.c,v
retrieving revision 1.7
diff -p -c -3 -r1.7 fstat.c
*** src/libc/posix/sys/stat/fstat.c     2001/09/25 01:00:52     1.7
--- src/libc/posix/sys/stat/fstat.c     2001/11/02 14:36:34
***************
*** 106,111 ****
--- 106,112 ----
  #include <dos.h>
  #include <sys/types.h>
  #include <sys/stat.h>
+ #include <libc/fd_props.h>
  
  #include <dpmi.h>
  #include <go32.h>
*************** fstat_assist(int fhandle, struct stat *s
*** 434,440 ****
    stat_buf->st_gid = getgid();
    stat_buf->st_nlink = 1;
  #ifndef  NO_ST_BLKSIZE
!   stat_buf->st_blksize = _go32_info_block.size_of_transfer_buffer;
  #endif
  
    /* If SFT entry for our handle is required and available, we will use
it.  */
--- 435,451 ----
    stat_buf->st_gid = getgid();
    stat_buf->st_nlink = 1;
  #ifndef  NO_ST_BLKSIZE
!   if (__get_fd_name(fhandle))
!     {
!       const char *filename = __get_fd_name(fhandle);
!       stat_buf->st_blksize = _get_cached_blksize(filename);
!     }
!   else
!     {
!       /* Fall back on transfer buffer size, if we can't determine file
name
!        * (which gives the drive letter and then the drive's cluster
size). */
!       stat_buf->st_blksize = _go32_info_block.size_of_transfer_buffer;
!     }
  #endif
  
    /* If SFT entry for our handle is required and available, we will use
it.  */
*************** fstat_assist(int fhandle, struct stat *s
*** 717,722 ****
--- 728,738 ----
  
              /* Additional time info for LFN platforms.  */
              set_fstat_times (fhandle, stat_buf);
+ 
+             /* Number of blocks */
+             stat_buf->st_blocks = _get_blkcnt(stat_buf->st_size,
+                                               stat_buf->st_blksize);
+ 
                return 0;
              }
  
*************** fstat_assist(int fhandle, struct stat *s
*** 856,861 ****
--- 872,881 ----
  
          /* Additional time info for LFN platforms.  */
          set_fstat_times (fhandle, stat_buf);
+ 
+         /* Number of blocks */
+         stat_buf->st_blocks = _get_blkcnt(stat_buf->st_size,
+                                           stat_buf->st_blksize);
          }
        return 0;
      }
*************** int main(int argc, char *argv[])
*** 936,941 ****
--- 956,964 ----
        _djstat_describe_lossage(stderr);
      }
  
+   if (!argc)
+     return EXIT_SUCCESS;
+ 
    /* Now call fstat() for each command-line argument. */
    while (++argv, --argc)
      {
*************** int main(int argc, char *argv[])
*** 956,961 ****
--- 979,987 ----
          fprintf (stderr, "\t\t\tTimes: %lu %lu\n",
                   (unsigned long)stat_buf.st_atime,
                   (unsigned long)stat_buf.st_ctime);
+         fprintf(stderr, "\t\t\tBlocks: %d %d\n",
+                 stat_buf.st_blksize,
+                 stat_buf.st_blocks);
            _djstat_describe_lossage(stderr);
          }
        else
*************** int main(int argc, char *argv[])
*** 965,971 ****
            _djstat_describe_lossage(stderr);
          }
      }
!   return 0;
  }
  
  #endif  /* TEST */
--- 991,997 ----
            _djstat_describe_lossage(stderr);
          }
      }
!   return EXIT_SUCCESS;
  }
  
  #endif  /* TEST */
Index: src/libc/posix/sys/stat/lstat.c
===================================================================
RCS file: /cvs/djgpp/djgpp/src/libc/posix/sys/stat/lstat.c,v
retrieving revision 1.7
diff -p -c -3 -r1.7 lstat.c
*** src/libc/posix/sys/stat/lstat.c     2001/10/17 05:08:39     1.7
--- src/libc/posix/sys/stat/lstat.c     2001/11/02 14:36:42
*************** stat_assist(const char *path, struct sta
*** 445,451 ****
    statbuf->st_gid     = getgid();
    statbuf->st_nlink   = 1;
  #ifndef  NO_ST_BLKSIZE
!   statbuf->st_blksize = _go32_info_block.size_of_transfer_buffer;
  #endif
  
    /* Make the path explicit.  This makes the rest of our job much
--- 445,451 ----
    statbuf->st_gid     = getgid();
    statbuf->st_nlink   = 1;
  #ifndef  NO_ST_BLKSIZE
!   statbuf->st_blksize = _get_cached_blksize(path);
  #endif
  
    /* Make the path explicit.  This makes the rest of our job much
*************** stat_assist(const char *path, struct sta
*** 868,873 ****
--- 868,876 ----
        statbuf->st_size  = (nfiles + extra) * sizeof(struct full_dirent);
      }
  
+   /* Number of blocks */
+   statbuf->st_blocks = _get_blkcnt(statbuf->st_size,
statbuf->st_blksize);
+ 
    return 0;
  }
  
*************** lstat(const char *path, struct stat *sta
*** 932,940 ****
  
  #ifdef  TEST
  
  unsigned short _djstat_flags = 0;
  
! void
  main(int argc, char *argv[])
  {
    struct stat stat_buf;
--- 935,945 ----
  
  #ifdef  TEST
  
+ #include <stdlib.h>
+ 
  unsigned short _djstat_flags = 0;
  
! int
  main(int argc, char *argv[])
  {
    struct stat stat_buf;
*************** main(int argc, char *argv[])
*** 943,949 ****
    if (argc < 2)
      {
        fprintf (stderr, "Usage: %s <_djstat_flags> <file...>\n",
argv[0]);
!       exit(0);
      }
  
    if (lstat(*argv, &stat_buf) != 0)
--- 948,954 ----
    if (argc < 2)
      {
        fprintf (stderr, "Usage: %s <_djstat_flags> <file...>\n",
argv[0]);
!       return (EXIT_FAILURE);
      }
  
    if (lstat(*argv, &stat_buf) != 0)
*************** main(int argc, char *argv[])
*** 968,973 ****
--- 973,981 ----
                    (long)stat_buf.st_size,
                    (unsigned long)stat_buf.st_mtime,
                    ctime(&stat_buf.st_mtime));
+         fprintf(stderr, "\t\t\tBlocks: %d %d\n",
+                 stat_buf.st_blksize,
+                 stat_buf.st_blocks);
            _djstat_describe_lossage(stderr);
          }
        else
*************** main(int argc, char *argv[])
*** 980,986 ****
        ++argv;
      }
  
!     exit (0);
  }
  
  #endif
--- 988,994 ----
        ++argv;
      }
  
!     return (EXIT_SUCCESS);
  }
  
  #endif
Index: src/libc/posix/sys/stat/xstat.c
===================================================================
RCS file: /cvs/djgpp/djgpp/src/libc/posix/sys/stat/xstat.c,v
retrieving revision 1.3
diff -p -c -3 -r1.3 xstat.c
*** src/libc/posix/sys/stat/xstat.c     2000/08/05 16:53:46     1.3
--- src/libc/posix/sys/stat/xstat.c     2001/11/02 14:36:50
***************
*** 19,24 ****
--- 19,27 ----
  #include <limits.h>
  #include <time.h>
  #include <errno.h>
+ #include <assert.h>
+ #include <ctype.h>
+ #include <sys/vfs.h>
  #include <dos.h>
  #include <dpmi.h>
  #include <libc/farptrgs.h>
*************** _getftime(int fhandle, unsigned int *dos
*** 116,121 ****
--- 119,176 ----
    *dos_ftime = ((unsigned int)regs.x.dx << 16) + (unsigned
int)regs.x.cx;
  
    return 0;
+ }
+ 
+ /* Cache the cluster size (aka block size) for each drive letter, so we
can
+  * populate the st_blksize of struct stat easily. The cluster size is
+  * measured in bytes.
+  */
+ 
+ /* Comment copied from DJGPP 2.03's src/libc/compat/mntent/mntent.c:
+  *
+  * There may be a maximum of 32 block devices.  Novell Netware indeed
+  * allows for 32 disks (A-Z plus 6 more characters from '[' to '\'')
+  */
+ static blksize_t cache_blksize[32];
+ static int cache_blksize_count = -1;
+ 
+ blksize_t
+ _get_cached_blksize (const char *path)
+ {  
+   char          fixed_path[PATH_MAX + 1];
+   struct statfs sbuf;
+   int           d; /* drive */  
+   static int    overmax_d = sizeof(cache_blksize) /
sizeof(cache_blksize[0]);
+ 
+   /* Force initialization in restarted programs (emacs).  */
+   if (cache_blksize_count != __bss_count)
+     {
+       cache_blksize_count = __bss_count;
+       memset(cache_blksize, 0, sizeof(cache_blksize));
+     }
+ 
+   _fixpath(path, fixed_path);
+   d = tolower(path[0]) - 'a';
+ 
+   if ((d < 0) || (d >= overmax_d))
+     {
+       errno = ENODEV;
+       return -1;
+     }
+ 
+   if (!cache_blksize[d])
+     {
+       /* No entry => retrieve cluster size */
+       if (statfs(path, &sbuf) != 0)
+       {
+         /* Failed, pass error through */
+         return -1;
+       }
+ 
+       cache_blksize[d] = sbuf.f_bsize;
+     }
+ 
+   return cache_blksize[d];
  }
  
  /* Invent an inode number for those files which don't have valid DOS
Index: src/libc/posix/sys/stat/stat.c
===================================================================
RCS file: /cvs/djgpp/djgpp/src/libc/posix/sys/stat/stat.c,v
retrieving revision 1.10
diff -p -c -3 -r1.10 stat.c
*** src/libc/posix/sys/stat/stat.c      2000/08/22 18:45:07     1.10
--- src/libc/posix/sys/stat/stat.c      2001/11/02 14:36:53
*************** stat(const char *path, struct stat *stat
*** 31,39 ****
  
  #ifdef  TEST
  
  unsigned short _djstat_flags = 0;
  
! void
  main(int argc, char *argv[])
  {
    struct stat stat_buf;
--- 31,41 ----
  
  #ifdef  TEST
  
+ #include <stdlib.h>
+ 
  unsigned short _djstat_flags = 0;
  
! int
  main(int argc, char *argv[])
  {
    struct stat stat_buf;
*************** main(int argc, char *argv[])
*** 42,48 ****
    if (argc < 2)
      {
        fprintf (stderr, "Usage: %s <_djstat_flags> <file...>\n",
argv[0]);
!       exit(0);
      }
  
    if (stat(*argv, &stat_buf) != 0)
--- 44,50 ----
    if (argc < 2)
      {
        fprintf (stderr, "Usage: %s <_djstat_flags> <file...>\n",
argv[0]);
!       return (EXIT_FAILURE);
      }
  
    if (stat(*argv, &stat_buf) != 0)
*************** main(int argc, char *argv[])
*** 67,72 ****
--- 69,77 ----
                    (long)stat_buf.st_size,
                    (unsigned long)stat_buf.st_mtime,
                    ctime(&stat_buf.st_mtime));
+         fprintf(stderr, "\t\t\tBlocks: %d %d\n",
+                 stat_buf.st_blksize,
+                 stat_buf.st_blocks);
            _djstat_describe_lossage(stderr);
          }
        else
*************** main(int argc, char *argv[])
*** 79,85 ****
        ++argv;
      }
  
!     exit (0);
  }
  
  #endif
--- 84,90 ----
        ++argv;
      }
  
!     return (EXIT_SUCCESS);
  }
  
  #endif
Index: src/libc/posix/sys/stat/xstat.h
===================================================================
RCS file: /cvs/djgpp/djgpp/src/libc/posix/sys/stat/xstat.h,v
retrieving revision 1.4
diff -p -c -3 -r1.4 xstat.h
*** src/libc/posix/sys/stat/xstat.h     2000/08/05 16:53:46     1.4
--- src/libc/posix/sys/stat/xstat.h     2001/11/02 14:36:57
*************** extern unsigned short   _djstat_fail_bit
*** 56,61 ****
--- 56,76 ----
  
  extern unsigned short   _djstat_flags;
  
+ static inline blkcnt_t _get_blkcnt (off_t _st_size, blksize_t
_st_blksize)
+ {
+   blkcnt_t _blkcnt;
+ 
+   if (!_st_size)
+     return 0;
+ 
+   /* Round up to the nearest block. */
+   _blkcnt = _st_size / _st_blksize;
+   if ((_blkcnt * _st_blksize) < _st_size)
+     _blkcnt++;
+ 
+   return _blkcnt;
+ }
+ 
  extern time_t           _file_time_stamp(unsigned int);
  extern ino_t            _invent_inode(const char *, unsigned, unsigned
long);
  extern unsigned short   _get_magic(const char *, int);
*************** extern long             __filelength(int
*** 68,72 ****
--- 83,88 ----
  extern int              _is_remote_handle(int);
  extern void             _djstat_describe_lossage(FILE *);
  extern int              _getftime(int, unsigned int *);
+ extern blksize_t        _get_cached_blksize (const char *path);
  
  #endif  /* __XSTAT_H */
Index: src/libc/posix/sys/stat/lstat.txh
===================================================================
RCS file: /cvs/djgpp/djgpp/src/libc/posix/sys/stat/lstat.txh,v
retrieving revision 1.3
diff -p -c -3 -r1.3 lstat.txh
*** src/libc/posix/sys/stat/lstat.txh   2001/08/01 10:31:02     1.3
--- src/libc/posix/sys/stat/lstat.txh   2001/11/02 14:37:06
*************** it in @var{sbuf}, which has this structu
*** 14,30 ****
  
  @smallexample
  struct  stat @{
!         time_t   st_atime;     /* time of last access       */
!         time_t   st_ctime;     /* time of file's creation   */
!         dev_t    st_dev;       /* The drive number (0 = a:) */
!         gid_t    st_gid;       /* what getgid() returns */
!         ino_t    st_ino;       /* starting cluster or unique identifier
*/
!         mode_t   st_mode;      /* file mode - S_IF* and S_IRUSR/S_IWUSR
*/
!         time_t   st_mtime;     /* time that the file was last written */
!         nlink_t  st_nlink;     /* 2 + number of subdirs, or 1 for files
*/
!         off_t    st_size;      /* size of file in bytes */
!         off_t    st_blksize;   /* the size of transfer buffer */
!         uid_t    st_uid;       /* what getuid() returns */
  @};
  @end smallexample
  
--- 14,31 ----
  
  @smallexample
  struct  stat @{
!         time_t    st_atime;     /* time of last access       */
!         time_t    st_ctime;     /* time of file's creation   */
!         dev_t     st_dev;       /* The drive number (0 = a:) */
!         gid_t     st_gid;       /* what getgid() returns */
!         ino_t     st_ino;       /* starting cluster or unique identifier
*/
!         mode_t    st_mode;      /* file mode - S_IF* and S_IRUSR/S_IWUSR
*/
!         time_t    st_mtime;     /* time that the file was last written
*/
!         nlink_t   st_nlink;     /* 2 + number of subdirs, or 1 for files
*/
!         off_t     st_size;      /* size of file in bytes */
!       blksize_t st_blksize;   /* block size in bytes*/
!         uid_t     st_uid;       /* what getuid() returns */
!       blkcnt_t  st_blocks;    /* blocks used by file */
  @};
  @end smallexample
  
Index: src/libc/posix/sys/stat/stat.txh
===================================================================
RCS file: /cvs/djgpp/djgpp/src/libc/posix/sys/stat/stat.txh,v
retrieving revision 1.8
diff -p -c -3 -r1.8 stat.txh
*** src/libc/posix/sys/stat/stat.txh    2001/08/01 10:31:02     1.8
--- src/libc/posix/sys/stat/stat.txh    2001/11/02 14:37:12
*************** it in @var{sbuf}, which has this structu
*** 14,30 ****
  
  @smallexample
  struct  stat @{
!         time_t   st_atime;     /* time of last access       */
!         time_t   st_ctime;     /* time of file's creation   */
!         dev_t    st_dev;       /* The drive number (0 = a:) */
!         gid_t    st_gid;       /* what getgid() returns */
!         ino_t    st_ino;       /* starting cluster or unique identifier
*/
!         mode_t   st_mode;      /* file mode - S_IF* and S_IRUSR/S_IWUSR
*/
!         time_t   st_mtime;     /* time that the file was last written */
!         nlink_t  st_nlink;     /* 2 + number of subdirs, or 1 for files
*/
!         off_t    st_size;      /* size of file in bytes */
!         off_t    st_blksize;   /* the size of transfer buffer */
!         uid_t    st_uid;       /* what getuid() returns */
  @};
  @end smallexample
  
--- 14,31 ----
  
  @smallexample
  struct  stat @{
!         time_t    st_atime;     /* time of last access       */
!         time_t    st_ctime;     /* time of file's creation   */
!         dev_t     st_dev;       /* The drive number (0 = a:) */
!         gid_t     st_gid;       /* what getgid() returns */
!         ino_t     st_ino;       /* starting cluster or unique identifier
*/
!         mode_t    st_mode;      /* file mode - S_IF* and S_IRUSR/S_IWUSR
*/
!         time_t    st_mtime;     /* time that the file was last written
*/
!         nlink_t   st_nlink;     /* 2 + number of subdirs, or 1 for files
*/
!         off_t     st_size;      /* size of file in bytes */
!         blksize_t st_blksize;   /* block size in bytes*/
!         uid_t     st_uid;       /* what getuid() returns */
!         blkcnt_t  st_blocks;    /* blocks used by file */
  @};
  @end smallexample
  
Index: src/docs/kb/wc204.txi
===================================================================
RCS file: /cvs/djgpp/djgpp/src/docs/kb/wc204.txi,v
retrieving revision 1.100
diff -p -c -3 -r1.100 wc204.txi
*** src/docs/kb/wc204.txi       2001/10/17 05:19:49     1.100
--- src/docs/kb/wc204.txi       2001/11/02 14:37:42
*************** case lowering.  This replaces the functi
*** 675,677 ****
--- 675,686 ----
  which is hopelessly buggy on Windows 2000 and XP.  New function used in
  @file{srchpath.c}, @file{readdir.c}, @file{glob.c}, @file{fixpath.c},
  @file{lstat.c} and @file{getcwd.c}.
+ 
+ @findex stat{, and block size}
+ @findex lstat{, and block size}
+ @findex fstat AT r{, and block size}
+ The functions @code{stat}, @code{lstat} and @code{fstat} now fill
+ the @code{st_blksize} member of @code{struct stat} with the correct
block
+ size for the device where the file is located. @code{struct stat} now
+ has a @code{st_blocks} member, which contains the number of blocks
+ occupied by the file.