[PATCH 0/3] fiemap patches (RFC/testing)

[Eric Sandeen <sandeen@redhat.com>]

Since Akira would like to use the fiemap ioctl for defrag, I thought I should
put what I have so far out on the list, at least.  This could go in the unstable
part of the tree if you like, though I need to do more testing etc before it's 
really ready to go.

Also, below is a quick test application I was using with the ioctl.

-Eric


#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <fcntl.h>
#include <string.h>
#include <sys/ioctl.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <sys/vfs.h>
#include <ext2fs/ext2_types.h>

/*************************************************/
/* All this should come from fiemap.h eventually */
struct fiemap_extent {
	__u64 fe_offset; /* offset in bytes for the start of the extent */
	__u64 fe_length; /* length in bytes for the extent */
	__u32 fe_flags;  /* returned FIEMAP_EXTENT_* flags for the extent */
	__u32 fe_lun;    /* logical device number for extent (starting at 0) */
};

struct fiemap {
	__u64 fm_start;	 /* logical starting byte offset (in/out) */
	__u64 fm_length;	/* logical length of map (in/out) */
	__u32 fm_flags;	 /* FIEMAP_FLAG_* flags for request (in/out) */
	__u32 fm_extent_count;  /* number of extents in fm_extents (in/out) */
	__u64 fm_end_offset;    /* logical offset of end of mapping in last ioctl */
	struct fiemap_extent fm_extents[0];
};

#define FIEMAP_FLAG_SYNC	0x00000001 /* sync file data before map */
#define FIEMAP_FLAG_HSM_READ    0x00000002 /* get data from HSM before map */
#define FIEMAP_FLAG_NUM_EXTENTS 0x00000004 /* return only number of extents */
#define FIEMAP_FLAG_INCOMPAT    0xff000000 /* error for unknown flags in here */
#define FIEMAP_FLAG_LUN_OFFSET  0x01000000 /* use lun offsets, instead of
					    * logical file offsets */

#define FIEMAP_EXTENT_HOLE      0x00000001 /* has no data or space allocation */
#define FIEMAP_EXTENT_UNWRITTEN 0x00000002 /* space allocated, but no data */
#define FIEMAP_EXTENT_UNMAPPED  0x00000004 /* has data but no space allocation*/
#define FIEMAP_EXTENT_ERROR     0x00000008 /* mapping error, errno in fe_start*/
#define FIEMAP_EXTENT_NO_DIRECT 0x00000010 /* cannot access data directly */
#define FIEMAP_EXTENT_LAST      0x00000020 /* last extent in the file */
#define FIEMAP_EXTENT_DELALLOC  0x00000040 /* has data but not yet written,
					      must have EXTENT_UNKNOWN set */
#define FIEMAP_EXTENT_SECONDARY 0x00000080 /* data (also) in secondary storage,
					      not in primary if EXTENT_UNKNOWN*/
#define FIEMAP_EXTENT_EOF       0x00000100 /* if fm_start+fm_len is beyond EOF*/

#define FIGETBSZ	_IO(0x00, 2)    /* get the block size used for bmap */
#define EXT4_IOC_FIEMAP _IOWR('f', 10, struct fiemap) /* get file extent info*/

/* End of what should be coming from fiemap.h */
/**********************************************/


void usage(void)
{
	printf("Usage: fiemap [-vrSCL] [-s start] [-l length] [-c buf count] [-m max] filename\n");
	printf("  -v        : verbose mode\n");
	printf("  -r        : raw output: print raw ioctl structure values\n");
	printf("  -S        : set FIEMAP_FLAG_SYNC to sync before mapping\n");
	printf("  -C        : set FIEMAP_FLAG_NUM_EXTENTS to only get extent count, not mapping\n");
	printf("  -L        : set FIEMAP_FLAG_LUN_OFFSET to report extents in lun order\n");
	printf("  -s start  : start of mapping in bytes (default 0)\n");
	printf("  -l length : length of mapping in bytes (default to end of file)\n");
	printf("  -c count  : count of extents in ioctl input structure (default 32)\n");
	printf("  -m max    : max nr of ioctls to call before exit (default 512)\n");
	exit(EXIT_FAILURE);
}

#define EXABYTES(x)     ((long long)(x) << 60)
#define PETABYTES(x)    ((long long)(x) << 50)
#define TERABYTES(x)    ((long long)(x) << 40)
#define GIGABYTES(x)    ((long long)(x) << 30)
#define MEGABYTES(x)    ((long long)(x) << 20)
#define KILOBYTES(x)    ((long long)(x) << 10)

long long
cvtnum(char *s)
{
	long long	i;
	char		*sp;
	int		c;

	i = strtoll(s, &sp, 0);
	if (i == 0 && sp == s)
		return -1LL;
	if (*sp == '\0')
		return i;
	if (sp[1] != '\0')
		return -1LL;

	c = tolower(*sp);
	switch (c) {
	case 'k':
		return KILOBYTES(i);
	case 'm':
		return MEGABYTES(i);
	case 'g':
		return GIGABYTES(i);
	case 't':
		return TERABYTES(i);
	case 'p':
		return PETABYTES(i);
	case 'e':
		return  EXABYTES(i);
	}

	return -1LL;
}

void show_extents_table(struct fiemap *fiemap, int blocksize, int start_extent, int *is_last)
{
	unsigned int	i;
	__u64		lstart;

	lstart = fiemap->fm_start;
	for (i = 0; i < fiemap->fm_extent_count; i++) {
		__u64	length = fiemap->fm_extents[i].fe_length;
		__u64	phys = fiemap->fm_extents[i].fe_offset;
		int	flags = fiemap->fm_extents[i].fe_flags;

		printf("ext: %3u logical: [%8llu..%8llu] phys: %8llu..%8llu flags: 0x%03X tot: %llu\n",
			i + start_extent,
			lstart, lstart + length - 1,
			(phys / blocksize),
			(flags & FIEMAP_EXTENT_HOLE) ? 0 : (phys + length - 1) / blocksize,
			flags,
			(length / blocksize));

		lstart += length;
		if (fiemap->fm_extents[i].fe_flags & FIEMAP_EXTENT_LAST) {
			*is_last = 1;
			return; /* XXX should we?  or should look for exents filled in past last? */
		}
	}
}

void show_extents_raw(struct fiemap *fiemap, int start_extent, int *is_last)
{
	unsigned int	i;

	for (i = 0; i < fiemap->fm_extent_count; i++) {
		printf("\tExtent %3u: start: %10lld length: %10lld flags 0x%03X lun %3u\n",
			i + start_extent,
			fiemap->fm_extents[i].fe_offset,
			fiemap->fm_extents[i].fe_length,
			fiemap->fm_extents[i].fe_flags,
			fiemap->fm_extents[i].fe_lun);

		if (fiemap->fm_extents[i].fe_flags & FIEMAP_EXTENT_LAST) {
			*is_last = 1;
			return; /* XXX should we?  or should look for exents filled in past last? */
		}
	}
}

int main(int argc, char**argv)
{
	int	blocksize = 0;	/* filesystem blocksize */
	uint	count = 32;	/* extent count */
	int	countonly = 0;	/* return only extent count? */
	int	fd;		/* file descriptor */
	int	last = 0;	/* last extent found */
	int	lunwise = 0;	/* return extents lun-wise */
	int	maxioctls = 512; /* max ioctls to try */
	int	opt;
	int	rc;
	int	raw = 0;	/* raw output format */
	int	sync = 0;	/* sync file before mapping */
	int	verbose = 0;	/* verbose output */
	char	*fname;		/* filename to map */
	char	*fiebuf;	/* fiemap buffer / ioctl argument */
	__u64	lstart = 0;	/* logical input mapping start */
	__u64	llength = ~0ULL;/* logical input mapping length */
	uint	start_ext = 0;	/* starting extent nr. for this batch */
	struct fiemap *fiemap;


	while ((opt = getopt(argc, argv, "s:l:c:m:rSCLv")) != -1) {
		switch(opt) {
		/* logical mapping offset */
		case 's':
			lstart = cvtnum(optarg);
			break;
		/* logical mapping length */
		case 'l':
			llength = cvtnum(optarg);
			break;
		/* count of extent buffers to send */
		case 'c':
			count = atoi(optarg);
			break;
		/* max nr. of ioctls to try (safety net) */
		case 'm':
			maxioctls = atoi(optarg);
			break;
		/* raw format output */
		case 'r':
			raw++;
			break;
		/* sync file before mapping */
		case 'S':
			sync++;
			break;
		/* count extents only, no details */
		case 'C':
			countonly++;
			break;
		/* return extents in lun order */
		case 'L':
			lunwise++;
			break;
		/* be verbose */
		case 'v':
			verbose++;
			break;
		default:
			usage();
		}
	}

	fname = argv[optind++];
	if (!fname)
		usage();

	/* The whole buffer, extent maps and all */
	fiebuf = malloc(sizeof (struct fiemap) + (count * sizeof(struct fiemap_extent)));
	if (!fiebuf) {
		perror("Could not allocate fiemap buffers");
		exit(1);
	}

	/* Just the header */
	fiemap = (struct fiemap *)fiebuf;

	fiemap->fm_start = lstart;
	fiemap->fm_length = llength;
	fiemap->fm_flags = 0;
	if (sync)
		fiemap->fm_flags |= FIEMAP_FLAG_SYNC;
	if (countonly)
		fiemap->fm_flags |= FIEMAP_FLAG_NUM_EXTENTS;
	if (lunwise)
		fiemap->fm_flags |= FIEMAP_FLAG_LUN_OFFSET;

	fiemap->fm_extent_count = count;
	fiemap->fm_end_offset = 0;	/* output only */

	fd = open(fname, O_RDONLY);
	if (fd < 0) {
		perror("Can't open file");
		exit(1);
	}

	if (ioctl(fd, FIGETBSZ, &blocksize) < 0) {
		perror("Can't get block size");
		close(fd);
		return;
	}

	do {
		if (verbose)
			printf("Input:  start %llu length %llu flags 0x%X count %u end_offset %lld\n",
				fiemap->fm_start, fiemap->fm_length,
				fiemap->fm_flags, fiemap->fm_extent_count,
				fiemap->fm_end_offset);

		rc = ioctl(fd, EXT4_IOC_FIEMAP, (unsigned long)fiemap);
		if (rc < 0) {
			perror("FIEMAP ioctl failed");
			close(fd);
			exit(1);
		}

		if (verbose)
			printf("Output: start %llu length %llu flags 0x%X count %u end_offset %lld\n",
				fiemap->fm_start, fiemap->fm_length,
				fiemap->fm_flags, fiemap->fm_extent_count,
				fiemap->fm_end_offset);
		if (raw)
			show_extents_raw(fiemap, start_ext, &last);
		else
			show_extents_table(fiemap, blocksize, start_ext, &last);

		start_ext += fiemap->fm_extent_count;

		/* Did we finish up the last of the reqest? */
		if (fiemap->fm_length >= llength)
			break;
		/* Set up the next call arguments */
		fiemap->fm_start += fiemap->fm_length;
		llength -= fiemap->fm_length;
		fiemap->fm_length = llength;
		fiemap->fm_extent_count = count;

		maxioctls--;

	} while (!last && maxioctls > 0);

	close(fd);

	return 0;
}