summaryrefslogtreecommitdiff
path: root/usr/src/boot/libsa/bzipfs.c
diff options
context:
space:
mode:
Diffstat (limited to 'usr/src/boot/libsa/bzipfs.c')
-rw-r--r--usr/src/boot/libsa/bzipfs.c404
1 files changed, 404 insertions, 0 deletions
diff --git a/usr/src/boot/libsa/bzipfs.c b/usr/src/boot/libsa/bzipfs.c
new file mode 100644
index 0000000000..1c2cc39904
--- /dev/null
+++ b/usr/src/boot/libsa/bzipfs.c
@@ -0,0 +1,404 @@
+/*
+ * Copyright (c) 1998 Michael Smith.
+ * Copyright (c) 2000 Maxim Sobolev
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ * 1. Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright
+ * notice, this list of conditions and the following disclaimer in the
+ * documentation and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
+ * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
+ * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+ * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
+ * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
+ * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
+ * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
+ * SUCH DAMAGE.
+ */
+
+#include <sys/cdefs.h>
+
+#ifndef REGRESSION
+#include "stand.h"
+#else
+#include <stdlib.h>
+#include <sys/errno.h>
+#include <sys/fcntl.h>
+#include <sys/types.h>
+#include <sys/unistd.h>
+
+struct open_file {
+ int f_flags; /* see F_* below */
+ void *f_fsdata; /* file system specific data */
+};
+#define F_READ 0x0001 /* file opened for reading */
+#define EOFFSET (ELAST + 8) /* relative seek not supported */
+#define panic(x, y) abort()
+
+static inline uint_t
+min(uint_t a, uint_t b)
+{
+ return (a < b ? a : b);
+}
+#endif
+
+#include <sys/stat.h>
+#include <string.h>
+#include <bzlib.h>
+
+#define BZ_BUFSIZE 2048 /* XXX larger? */
+
+struct bz_file
+{
+ int bzf_rawfd;
+ bz_stream bzf_bzstream;
+ char bzf_buf[BZ_BUFSIZE];
+ int bzf_endseen;
+};
+
+static int bzf_fill(struct bz_file *);
+static int bzf_open(const char *, struct open_file *);
+static int bzf_close(struct open_file *);
+static int bzf_read(struct open_file *, void *, size_t, size_t *);
+static off_t bzf_seek(struct open_file *, off_t, int);
+static int bzf_stat(struct open_file *, struct stat *);
+
+#ifndef REGRESSION
+struct fs_ops bzipfs_fsops = {
+ .fs_name = "bzip",
+ .fo_open = bzf_open,
+ .fo_close = bzf_close,
+ .fo_read = bzf_read,
+ .fo_write = null_write,
+ .fo_seek = bzf_seek,
+ .fo_stat = bzf_stat,
+ .fo_readdir = null_readdir
+};
+#endif
+
+static int
+bzf_fill(struct bz_file *bzf)
+{
+ int result;
+ int req;
+
+ req = BZ_BUFSIZE - bzf->bzf_bzstream.avail_in;
+ result = 0;
+
+ /* If we need more */
+ if (req > 0) {
+ /* move old data to bottom of buffer */
+ if (req < BZ_BUFSIZE) {
+ bcopy(bzf->bzf_buf + req, bzf->bzf_buf,
+ BZ_BUFSIZE - req);
+ }
+
+ /* read to fill buffer and update availibility data */
+ result = read(bzf->bzf_rawfd,
+ bzf->bzf_buf + bzf->bzf_bzstream.avail_in, req);
+ bzf->bzf_bzstream.next_in = bzf->bzf_buf;
+ if (result >= 0)
+ bzf->bzf_bzstream.avail_in += result;
+ }
+ return (result);
+}
+
+/*
+ * Adapted from get_byte/check_header in libz
+ *
+ * Returns 0 if the header is OK, nonzero if not.
+ */
+static int
+get_byte(struct bz_file *bzf)
+{
+ if ((bzf->bzf_bzstream.avail_in == 0) && (bzf_fill(bzf) == -1))
+ return (-1);
+ bzf->bzf_bzstream.avail_in--;
+ return (*(bzf->bzf_bzstream.next_in)++);
+}
+
+static int bz_magic[3] = {'B', 'Z', 'h'}; /* bzip2 magic header */
+
+static int
+check_header(struct bz_file *bzf)
+{
+ unsigned int len;
+ int c;
+
+ /* Check the bzip2 magic header */
+ for (len = 0; len < 3; len++) {
+ c = get_byte(bzf);
+ if (c != bz_magic[len]) {
+ return (1);
+ }
+ }
+ /* Check that the block size is valid */
+ c = get_byte(bzf);
+ if (c < '1' || c > '9')
+ return (1);
+
+ /* Put back bytes that we've took from the input stream */
+ bzf->bzf_bzstream.next_in -= 4;
+ bzf->bzf_bzstream.avail_in += 4;
+
+ return (0);
+}
+
+static int
+bzf_open(const char *fname, struct open_file *f)
+{
+ static char *bzfname;
+ int rawfd;
+ struct bz_file *bzf;
+ char *cp;
+ int error;
+ struct stat sb;
+
+ /* Have to be in "just read it" mode */
+ if (f->f_flags != F_READ)
+ return (EPERM);
+
+ /* If the name already ends in .gz or .bz2, ignore it */
+ if ((cp = strrchr(fname, '.')) &&
+ ((strcmp(cp, ".gz") == 0) ||
+ (strcmp(cp, ".bz2") == 0) ||
+ (strcmp(cp, ".split") == 0)))
+ return (ENOENT);
+
+ /* Construct new name */
+ bzfname = malloc(strlen(fname) + 5);
+ if (bzfname == NULL)
+ return (ENOMEM);
+ sprintf(bzfname, "%s.bz2", fname);
+
+ /* Try to open the compressed datafile */
+ rawfd = open(bzfname, O_RDONLY);
+ free(bzfname);
+ if (rawfd == -1)
+ return (ENOENT);
+
+ if (fstat(rawfd, &sb) < 0) {
+ printf("bzf_open: stat failed\n");
+ close(rawfd);
+ return (ENOENT);
+ }
+ if (!S_ISREG(sb.st_mode)) {
+ printf("bzf_open: not a file\n");
+ close(rawfd);
+ return (EISDIR); /* best guess */
+ }
+
+ /* Allocate a bz_file structure, populate it */
+ bzf = malloc(sizeof (struct bz_file));
+ if (bzf == NULL)
+ return (ENOMEM);
+ bzero(bzf, sizeof (struct bz_file));
+ bzf->bzf_rawfd = rawfd;
+
+ /* Verify that the file is bzipped */
+ if (check_header(bzf)) {
+ close(bzf->bzf_rawfd);
+ free(bzf);
+ return (EFTYPE);
+ }
+
+ /* Initialise the inflation engine */
+ error = BZ2_bzDecompressInit(&(bzf->bzf_bzstream), 0, 1);
+ if (error != BZ_OK) {
+ printf("bzf_open: BZ2_bzDecompressInit returned %d\n", error);
+ close(bzf->bzf_rawfd);
+ free(bzf);
+ return (EIO);
+ }
+
+ /* Looks OK, we'll take it */
+ f->f_fsdata = bzf;
+ return (0);
+}
+
+static int
+bzf_close(struct open_file *f)
+{
+ struct bz_file *bzf = (struct bz_file *)f->f_fsdata;
+
+ BZ2_bzDecompressEnd(&(bzf->bzf_bzstream));
+ close(bzf->bzf_rawfd);
+ free(bzf);
+ return (0);
+}
+
+static int
+bzf_read(struct open_file *f, void *buf, size_t size, size_t *resid)
+{
+ struct bz_file *bzf = (struct bz_file *)f->f_fsdata;
+ int error;
+
+ bzf->bzf_bzstream.next_out = buf; /* where and how much */
+ bzf->bzf_bzstream.avail_out = size;
+
+ while (bzf->bzf_bzstream.avail_out && bzf->bzf_endseen == 0) {
+ if ((bzf->bzf_bzstream.avail_in == 0) &&
+ (bzf_fill(bzf) == -1)) {
+ printf("bzf_read: fill error\n");
+ return (EIO);
+ }
+ if (bzf->bzf_bzstream.avail_in == 0) {
+ /* oops, unexpected EOF */
+ printf("bzf_read: unexpected EOF\n");
+ if (bzf->bzf_bzstream.avail_out == size)
+ return (EIO);
+ break;
+ }
+
+ /* decompression pass */
+ error = BZ2_bzDecompress(&bzf->bzf_bzstream);
+ if (error == BZ_STREAM_END) { /* EOF, all done */
+ bzf->bzf_endseen = 1;
+ break;
+ }
+ if (error != BZ_OK) { /* argh, decompression error */
+ printf("bzf_read: BZ2_bzDecompress returned %d\n",
+ error);
+ return (EIO);
+ }
+ }
+ if (resid != NULL)
+ *resid = bzf->bzf_bzstream.avail_out;
+ return (0);
+}
+
+static int
+bzf_rewind(struct open_file *f)
+{
+ struct bz_file *bzf = (struct bz_file *)f->f_fsdata;
+ struct bz_file *bzf_tmp;
+
+ /*
+ * Since bzip2 does not have an equivalent inflateReset function a crude
+ * one needs to be provided. The functions all called in such a way that
+ * at any time an error occurs a roll back can be done (effectively
+ * making this rewind 'atomic', either the reset occurs successfully
+ * or not at all, with no 'undefined' state happening).
+ */
+
+ /* Allocate a bz_file structure, populate it */
+ bzf_tmp = malloc(sizeof (struct bz_file));
+ if (bzf_tmp == NULL)
+ return (-1);
+ bzero(bzf_tmp, sizeof (struct bz_file));
+ bzf_tmp->bzf_rawfd = bzf->bzf_rawfd;
+
+ /* Initialise the inflation engine */
+ if (BZ2_bzDecompressInit(&(bzf_tmp->bzf_bzstream), 0, 1) != BZ_OK) {
+ free(bzf_tmp);
+ return (-1);
+ }
+
+ /* Seek back to the beginning of the file */
+ if (lseek(bzf->bzf_rawfd, 0, SEEK_SET) == -1) {
+ BZ2_bzDecompressEnd(&(bzf_tmp->bzf_bzstream));
+ free(bzf_tmp);
+ return (-1);
+ }
+
+ /* Free old bz_file data */
+ BZ2_bzDecompressEnd(&(bzf->bzf_bzstream));
+ free(bzf);
+
+ /* Use the new bz_file data */
+ f->f_fsdata = bzf_tmp;
+
+ return (0);
+}
+
+static off_t
+bzf_seek(struct open_file *f, off_t offset, int where)
+{
+ struct bz_file *bzf = (struct bz_file *)f->f_fsdata;
+ off_t target;
+ char discard[16];
+
+ switch (where) {
+ case SEEK_SET:
+ target = offset;
+ break;
+ case SEEK_CUR:
+ target = offset + bzf->bzf_bzstream.total_out_lo32;
+ break;
+ default:
+ errno = EINVAL;
+ return (-1);
+ }
+
+ /* Can we get there from here? */
+ if (target < bzf->bzf_bzstream.total_out_lo32 && bzf_rewind(f) != 0) {
+ errno = EOFFSET;
+ return (-1);
+ }
+
+ /* if bzf_rewind was called then bzf has changed */
+ bzf = (struct bz_file *)f->f_fsdata;
+
+ /* skip forwards if required */
+ while (target > bzf->bzf_bzstream.total_out_lo32) {
+ errno = bzf_read(f, discard, min(sizeof (discard),
+ target - bzf->bzf_bzstream.total_out_lo32), NULL);
+ if (errno)
+ return (-1);
+ }
+ /* This is where we are (be honest if we overshot) */
+ return (bzf->bzf_bzstream.total_out_lo32);
+}
+
+static int
+bzf_stat(struct open_file *f, struct stat *sb)
+{
+ struct bz_file *bzf = (struct bz_file *)f->f_fsdata;
+ int result;
+
+ /* stat as normal, but indicate that size is unknown */
+ if ((result = fstat(bzf->bzf_rawfd, sb)) == 0)
+ sb->st_size = -1;
+ return (result);
+}
+
+void
+bz_internal_error(int errorcode)
+{
+ panic("bzipfs: critical error %d in bzip2 library occured",
+ errorcode);
+}
+
+#ifdef REGRESSION
+/* Small test case, open and decompress test.bz2 */
+int
+main()
+{
+ struct open_file f;
+ char buf[1024];
+ size_t resid;
+ int err;
+
+ memset(&f, '\0', sizeof (f));
+ f.f_flags = F_READ;
+ err = bzf_open("test", &f);
+ if (err != 0)
+ exit(1);
+ do {
+ err = bzf_read(&f, buf, sizeof (buf), &resid);
+ } while (err == 0 && resid != sizeof (buf));
+
+ if (err != 0)
+ exit(2);
+ exit(0);
+}
+#endif