/* * tar - `tape archiver', actually usable on any medium. * POSIX "ustar" compliant when extracting, and by default when creating. * this tar attempts to read and write multiple Tblock-byte blocks * at once to and from the filesystem, and does not copy blocks * around internally. */ #include <u.h>#include <libc.h>#include <thread.h>#include <fcall.h> /* for %M */#include <libString.h> /* * modified versions of those in libc.h; scans only the first arg for * keyletters and options. */#define TARGBEGIN {\ if (!argv0) argv0 = *argv; \ argv++, argc--;\ if (argv[0]) {\ char *_args, *_argt;\ Rune _argc;\ _args = &argv[0][0];\ _argc = 0;\ while(*_args && (_args += chartorune(&_argc, _args)))\ switch(_argc)#define TARGEND SET(_argt); USED(_argt);USED(_argc);USED(_args); \ argc--, argv++; } \ USED(argv); USED(argc); }#define TARGC() (_argc) #define ROUNDUP(a, b) (((a) + (b) - 1)/(b))#define BYTES2TBLKS(bytes) ROUNDUP(bytes, Tblock) /* read big-endian binary integers; args must be (uchar *) */#define G2BEBYTE(x) (((x)[0]<<8) | (x)[1])#define G3BEBYTE(x) (((x)[0]<<16) | ((x)[1]<<8) | (x)[2])#define G4BEBYTE(x) (((x)[0]<<24) | ((x)[1]<<16) | ((x)[2]<<8) | (x)[3])#define G8BEBYTE(x) (((vlong)G4BEBYTE(x)<<32) | (u32int)G4BEBYTE((x)+4)) typedef vlong Off;typedef char *(*Refill)(int ar, char *bufs, int justhdr); enum { Stdin, Stdout, Stderr };enum { Rd, Wr }; /* pipe fd-array indices */enum { Output, Input };enum { None, Toc, Xtract, Replace };enum { Alldata, Justnxthdr };enum { Tblock = 512, Namsiz = 100, Maxpfx = 155, /* from POSIX */ Maxname = Namsiz + 1 + Maxpfx, Binsize = 0x80, /* flag in size[0], from gnu: positive binary size */ Binnegsz = 0xff, /* flag in size[0]: negative binary size */ Nblock = 40, /* maximum blocksize */ Dblock = 20, /* default blocksize */ DEBUG = 0}; /* POSIX link flags */enum { LF_PLAIN1 = '\0', LF_PLAIN2 = '0', LF_LINK = '1', LF_SYMLINK1 = '2', LF_SYMLINK2 = 's', /* 4BSD used this */ LF_CHR = '3', LF_BLK = '4', LF_DIR = '5', LF_FIFO = '6', LF_CONTIG = '7' /* 'A' - 'Z' are reserved for custom implementations */}; #define islink(lf) (isreallink(lf) || issymlink(lf))#define isreallink(lf) ((lf) == LF_LINK)#define issymlink(lf) ((lf) == LF_SYMLINK1 || (lf) == LF_SYMLINK2) typedef struct Hdr { char name[Namsiz]; char mode[8]; char uid[8]; char gid[8]; char size[12]; char mtime[12]; char chksum[8]; char linkflag; char linkname[Namsiz]; /* rest are defined by POSIX's ustar format; see p1003.2b */ char magic[6]; /* "ustar" */ char version[2]; char uname[32]; char gname[32]; char devmajor[8]; char devminor[8]; char prefix[Maxpfx]; /* if non-null, path= prefix "/" name */ char pad[12];} Hdr; typedef struct Compress { char *comp; char *decomp; char *sfx[4];} Compress; static Compress comps[] = { "gzip", "gunzip", { ".tar.gz", ".tgz" }, /* default */ "compress", "uncompress", { ".tar.Z", ".tz" }, "bzip2", "bunzip2", { ".tar.bz", ".tbz", ".tar.bz2",".tbz2" },}; typedef struct Pushstate { int kid; int fd; /* original fd */ int rfd; /* replacement fd */ int input; int open;} Pushstate; #define OTHER(rdwr) (rdwr == Rd? Wr: Rd) /* static int debug; */static int verb;static int posix = 1;static int docreate;static int aruid;static int argid;static int relative = 1;static int settime;static int verbose;static int docompress;static int keepexisting;static Off blkoff; /* offset of the current archive block (not Tblock) */static Off nexthdr; static int nblock = Dblock;static char *usefile;static char origdir[Maxname*2];static Hdr *tpblk, *endblk;static Hdr *curblk; static voidusage(void){ fprint(2, "usage: %s {crtx}[PRTfgkmpuvz] [archive] file1 file2...\n", argv0); threadexitsall("usage");} /* compression */ static Compress *compmethod(char *name){ int i, nmlen = strlen(name), sfxlen; Compress *cp; for (cp = comps; cp < comps + nelem(comps); cp++) for (i = 0; i < nelem(cp->sfx) && cp->sfx[i]; i++) { sfxlen = strlen(cp->sfx[i]); if (nmlen > sfxlen && strcmp(cp->sfx[i], name + nmlen - sfxlen) == 0) return cp; } return docompress? comps: nil;} /* * push a filter, cmd, onto fd. if input, it's an input descriptor. * returns a descriptor to replace fd, or -1 on error. */static intpush(int fd, char *cmd, int input, Pushstate *ps){ int nfd, pifds[2], fds[3]; ps->open = 0; ps->fd = fd; ps->input = input; if (fd < 0 || pipe(pifds) < 0) return -1; if (input) fds[Stdout] = pifds[Wr]; else fds[Stdin] = pifds[Rd]; nfd = pifds[input? Rd: Wr]; fds[input? Stdin: Stdout] = dup(fd, -1); fds[Stderr] = dup(Stderr, -1); ps->kid = threadspawnl(fds, cmd, cmd, nil); if (ps->kid==-1) sysfatal("can't spawn %s: %r", cmd); ps->rfd = nfd; ps->open = 1; return nfd;} static char *pushclose(Pushstate *ps){ Waitmsg *wm; if (ps->fd < 0 || ps->rfd < 0 || !ps->open) return "not open"; close(ps->rfd); ps->rfd = -1; ps->open = 0; while ((wm = wait()) != nil && wm->pid != ps->kid) continue; return wm? wm->msg: nil;} /* * block-buffer management */ static voidinitblks(void){ free(tpblk); tpblk = malloc(Tblock * nblock); assert(tpblk != nil); endblk = tpblk + nblock;} /* * (re)fill block buffers from archive. `justhdr' means we don't care * about the data before the next header block. */static char *refill(int ar, char *bufs, int justhdr){ int i, n; unsigned bytes = Tblock * nblock; static int done, first = 1, seekable; if (done) return nil; if (first) seekable = seek(ar, 0, 1) >= 0; blkoff = seek(ar, 0, 1); /* note position for `tar r' */ /* try to size non-pipe input at first read */ if (first && usefile) { n = read(ar, bufs, bytes); if (n <= 0) sysfatal("error reading archive: %r"); i = n; if (i % Tblock != 0) { fprint(2, "%s: archive block size (%d) error\n", argv0, i); threadexitsall("blocksize"); } i /= Tblock; if (i != nblock) { nblock = i; fprint(2, "%s: blocking = %d\n", argv0, nblock); endblk = (Hdr *)bufs + nblock; bytes = n; } } else if (justhdr && seekable && nexthdr - seek(ar, 0, 1) >= bytes) { /* optimisation for huge archive members on seekable media */ if (seek(ar, bytes, 1) < 0) sysfatal("can't seek on archive: %r"); n = bytes; } else n = readn(ar, bufs, bytes); first = 0; if (n == 0) sysfatal("unexpected EOF reading archive"); else if (n < 0) sysfatal("error reading archive: %r"); else if (n%Tblock != 0) sysfatal("partial block read from archive"); if (n != bytes) { done = 1; memset(bufs + n, 0, bytes - n); } return bufs;} static Hdr *getblk(int ar, Refill rfp, int justhdr){ if (curblk == nil || curblk >= endblk) { /* input block exhausted? */ if (rfp != nil && (*rfp)(ar, (char *)tpblk, justhdr) == nil) return nil; curblk = tpblk; } return curblk++;} static Hdr *getblkrd(int ar, int justhdr){ return getblk(ar, refill, justhdr);} static Hdr *getblke(int ar){ return getblk(ar, nil, Alldata);} static Hdr *getblkz(int ar){ Hdr *hp = getblke(ar); if (hp != nil) memset(hp, 0, Tblock); return hp;} /* * how many block buffers are available, starting at the address * just returned by getblk*? */static intgothowmany(int max){ int n = endblk - (curblk - 1); return n > max? max: n;} /* * indicate that one is done with the last block obtained from getblke * and it is now available to be written into the archive. */static voidputlastblk(int ar){ unsigned bytes = Tblock * nblock; /* if writing end-of-archive, aid compression (good hygiene too) */ if (curblk < endblk) memset(curblk, 0, (char *)endblk - (char *)curblk); if (write(ar, tpblk, bytes) != bytes) sysfatal("error writing archive: %r");} static voidputblk(int ar){ if (curblk >= endblk) putlastblk(ar);} static voidputbackblk(int ar){ curblk--; USED(ar);} static voidputreadblks(int ar, int blks){ curblk += blks - 1; USED(ar);} static voidputblkmany(int ar, int blks){ curblk += blks - 1; putblk(ar);} /* * common routines */ /* * modifies hp->chksum but restores it; important for the last block of the * old archive when updating with `tar rf archive' */static longchksum(Hdr *hp){ int n = Tblock; long i = 0; uchar *cp = (uchar*)hp; char oldsum[sizeof hp->chksum]; memmove(oldsum, hp->chksum, sizeof oldsum); memset(hp->chksum, ' ', sizeof hp->chksum); while (n-- > 0) i += *cp++; memmove(hp->chksum, oldsum, sizeof oldsum); return i;} static intisustar(Hdr *hp){ return strcmp(hp->magic, "ustar") == 0;} /* * s is at most n bytes long, but need not be NUL-terminated. * if shorter than n bytes, all bytes after the first NUL must also * be NUL. */static intsstrnlen(char *s, int n){ return s[n - 1] != '\0'? n: strlen(s);} /* set fullname from header */static char *name(Hdr *hp){ int pfxlen, namlen; static char fullnamebuf[2 + Maxname + 1]; /* 2 at beginning for ./ on relative names */ char *fullname; fullname = fullnamebuf+2; namlen = sstrnlen(hp->name, sizeof hp->name); if (hp->prefix[0] == '\0' || !isustar(hp)) { /* old-style name? */ memmove(fullname, hp->name, namlen); fullname[namlen] = '\0'; return fullname; } /* name is in two pieces */ pfxlen = sstrnlen(hp->prefix, sizeof hp->prefix); memmove(fullname, hp->prefix, pfxlen); fullname[pfxlen] = '/'; memmove(fullname + pfxlen + 1, hp->name, namlen); fullname[pfxlen + 1 + namlen] = '\0'; return fullname;} static intisdir(Hdr *hp){ /* the mode test is ugly but sometimes necessary */ return hp->linkflag == LF_DIR || strrchr(name(hp), '\0')[-1] == '/' || (strtoul(hp->mode, nil, 8)&0170000) == 040000;} static inteotar(Hdr *hp){ return name(hp)[0] == '\0';} /*static uvlonggetbe(uchar *src, int size){ uvlong vl = 0; while (size-- > 0) { vl <<= 8; vl |= *src++; } return vl;} */ static voidputbe(uchar *dest, uvlong vl, int size){ for (dest += size; size-- > 0; vl >>= 8) *--dest = vl;} /* * return the nominal size from the header block, which is not always the * size in the archive (the archive size may be zero for some file types * regardless of the nominal size). * * gnu and freebsd tars are now recording vlongs as big-endian binary * with a flag in byte 0 to indicate this, which permits file sizes up to * 2^64-1 (actually 2^80-1 but our file sizes are vlongs) rather than 2^33-1. */static Offhdrsize(Hdr *hp){ uchar *p; if((uchar)hp->size[0] == Binnegsz) { fprint(2, "%s: %s: negative length, which is insane\n", argv0, name(hp)); return 0; } else if((uchar)hp->size[0] == Binsize) { p = (uchar *)hp->size + sizeof hp->size - 1 - sizeof(vlong); /* -1 for terminating space */Lines 1–500 of 1164. Load more
/* * tar - `tape archiver', actually usable on any medium. * POSIX "ustar" compliant when extracting, and by default when creating. * this tar attempts to read and write multiple Tblock-byte blocks * at once to and from the filesystem, and does not copy blocks * around internally. */ #include <u.h> #include <libc.h> #include <thread.h> #include <fcall.h> /* for %M */ #include <libString.h> /* * modified versions of those in libc.h; scans only the first arg for * keyletters and options. */ #define TARGBEGIN {\ if (!argv0) argv0 = *argv; \ argv++, argc--;\ if (argv[0]) {\ char *_args, *_argt;\ Rune _argc;\ _args = &argv[0][0];\ _argc = 0;\ while(*_args && (_args += chartorune(&_argc, _args)))\ switch(_argc) #define TARGEND SET(_argt); USED(_argt);USED(_argc);USED(_args); \ argc--, argv++; } \ USED(argv); USED(argc); } #define TARGC() (_argc) #define ROUNDUP(a, b) (((a) + (b) - 1)/(b)) #define BYTES2TBLKS(bytes) ROUNDUP(bytes, Tblock) /* read big-endian binary integers; args must be (uchar *) */ #define G2BEBYTE(x) (((x)[0]<<8) | (x)[1]) #define G3BEBYTE(x) (((x)[0]<<16) | ((x)[1]<<8) | (x)[2]) #define G4BEBYTE(x) (((x)[0]<<24) | ((x)[1]<<16) | ((x)[2]<<8) | (x)[3]) #define G8BEBYTE(x) (((vlong)G4BEBYTE(x)<<32) | (u32int)G4BEBYTE((x)+4)) typedef vlong Off; typedef char *(*Refill)(int ar, char *bufs, int justhdr); enum { Stdin, Stdout, Stderr }; enum { Rd, Wr }; /* pipe fd-array indices */ enum { Output, Input }; enum { None, Toc, Xtract, Replace }; enum { Alldata, Justnxthdr }; enum { Tblock = 512, Namsiz = 100, Maxpfx = 155, /* from POSIX */ Maxname = Namsiz + 1 + Maxpfx, Binsize = 0x80, /* flag in size[0], from gnu: positive binary size */ Binnegsz = 0xff, /* flag in size[0]: negative binary size */ Nblock = 40, /* maximum blocksize */ Dblock = 20, /* default blocksize */ DEBUG = 0 }; /* POSIX link flags */ enum { LF_PLAIN1 = '\0', LF_PLAIN2 = '0', LF_LINK = '1', LF_SYMLINK1 = '2', LF_SYMLINK2 = 's', /* 4BSD used this */ LF_CHR = '3', LF_BLK = '4', LF_DIR = '5', LF_FIFO = '6', LF_CONTIG = '7' /* 'A' - 'Z' are reserved for custom implementations */ }; #define islink(lf) (isreallink(lf) || issymlink(lf)) #define isreallink(lf) ((lf) == LF_LINK) #define issymlink(lf) ((lf) == LF_SYMLINK1 || (lf) == LF_SYMLINK2) typedef struct Hdr { char name[Namsiz]; char mode[8]; char uid[8]; char gid[8]; char size[12]; char mtime[12]; char chksum[8]; char linkflag; char linkname[Namsiz]; /* rest are defined by POSIX's ustar format; see p1003.2b */ char magic[6]; /* "ustar" */ char version[2]; char uname[32]; char gname[32]; char devmajor[8]; char devminor[8]; char prefix[Maxpfx]; /* if non-null, path= prefix "/" name */ char pad[12]; } Hdr; typedef struct Compress { char *comp; char *decomp; char *sfx[4]; } Compress; static Compress comps[] = { "gzip", "gunzip", { ".tar.gz", ".tgz" }, /* default */ "compress", "uncompress", { ".tar.Z", ".tz" }, "bzip2", "bunzip2", { ".tar.bz", ".tbz", ".tar.bz2",".tbz2" }, }; typedef struct Pushstate { int kid; int fd; /* original fd */ int rfd; /* replacement fd */ int input; int open; } Pushstate; #define OTHER(rdwr) (rdwr == Rd? Wr: Rd) /* static int debug; */ static int verb; static int posix = 1; static int docreate; static int aruid; static int argid; static int relative = 1; static int settime; static int verbose; static int docompress; static int keepexisting; static Off blkoff; /* offset of the current archive block (not Tblock) */ static Off nexthdr; static int nblock = Dblock; static char *usefile; static char origdir[Maxname*2]; static Hdr *tpblk, *endblk; static Hdr *curblk; static void usage(void) { fprint(2, "usage: %s {crtx}[PRTfgkmpuvz] [archive] file1 file2...\n", argv0); threadexitsall("usage"); } /* compression */ static Compress * compmethod(char *name) { int i, nmlen = strlen(name), sfxlen; Compress *cp; for (cp = comps; cp < comps + nelem(comps); cp++) for (i = 0; i < nelem(cp->sfx) && cp->sfx[i]; i++) { sfxlen = strlen(cp->sfx[i]); if (nmlen > sfxlen && strcmp(cp->sfx[i], name + nmlen - sfxlen) == 0) return cp; } return docompress? comps: nil; } /* * push a filter, cmd, onto fd. if input, it's an input descriptor. * returns a descriptor to replace fd, or -1 on error. */ static int push(int fd, char *cmd, int input, Pushstate *ps) { int nfd, pifds[2], fds[3]; ps->open = 0; ps->fd = fd; ps->input = input; if (fd < 0 || pipe(pifds) < 0) return -1; if (input) fds[Stdout] = pifds[Wr]; else fds[Stdin] = pifds[Rd]; nfd = pifds[input? Rd: Wr]; fds[input? Stdin: Stdout] = dup(fd, -1); fds[Stderr] = dup(Stderr, -1); ps->kid = threadspawnl(fds, cmd, cmd, nil); if (ps->kid==-1) sysfatal("can't spawn %s: %r", cmd); ps->rfd = nfd; ps->open = 1; return nfd; } static char * pushclose(Pushstate *ps) { Waitmsg *wm; if (ps->fd < 0 || ps->rfd < 0 || !ps->open) return "not open"; close(ps->rfd); ps->rfd = -1; ps->open = 0; while ((wm = wait()) != nil && wm->pid != ps->kid) continue; return wm? wm->msg: nil; } /* * block-buffer management */ static void initblks(void) { free(tpblk); tpblk = malloc(Tblock * nblock); assert(tpblk != nil); endblk = tpblk + nblock; } /* * (re)fill block buffers from archive. `justhdr' means we don't care * about the data before the next header block. */ static char * refill(int ar, char *bufs, int justhdr) { int i, n; unsigned bytes = Tblock * nblock; static int done, first = 1, seekable; if (done) return nil; if (first) seekable = seek(ar, 0, 1) >= 0; blkoff = seek(ar, 0, 1); /* note position for `tar r' */ /* try to size non-pipe input at first read */ if (first && usefile) { n = read(ar, bufs, bytes); if (n <= 0) sysfatal("error reading archive: %r"); i = n; if (i % Tblock != 0) { fprint(2, "%s: archive block size (%d) error\n", argv0, i); threadexitsall("blocksize"); } i /= Tblock; if (i != nblock) { nblock = i; fprint(2, "%s: blocking = %d\n", argv0, nblock); endblk = (Hdr *)bufs + nblock; bytes = n; } } else if (justhdr && seekable && nexthdr - seek(ar, 0, 1) >= bytes) { /* optimisation for huge archive members on seekable media */ if (seek(ar, bytes, 1) < 0) sysfatal("can't seek on archive: %r"); n = bytes; } else n = readn(ar, bufs, bytes); first = 0; if (n == 0) sysfatal("unexpected EOF reading archive"); else if (n < 0) sysfatal("error reading archive: %r"); else if (n%Tblock != 0) sysfatal("partial block read from archive"); if (n != bytes) { done = 1; memset(bufs + n, 0, bytes - n); } return bufs; } static Hdr * getblk(int ar, Refill rfp, int justhdr) { if (curblk == nil || curblk >= endblk) { /* input block exhausted? */ if (rfp != nil && (*rfp)(ar, (char *)tpblk, justhdr) == nil) return nil; curblk = tpblk; } return curblk++; } static Hdr * getblkrd(int ar, int justhdr) { return getblk(ar, refill, justhdr); } static Hdr * getblke(int ar) { return getblk(ar, nil, Alldata); } static Hdr * getblkz(int ar) { Hdr *hp = getblke(ar); if (hp != nil) memset(hp, 0, Tblock); return hp; } /* * how many block buffers are available, starting at the address * just returned by getblk*? */ static int gothowmany(int max) { int n = endblk - (curblk - 1); return n > max? max: n; } /* * indicate that one is done with the last block obtained from getblke * and it is now available to be written into the archive. */ static void putlastblk(int ar) { unsigned bytes = Tblock * nblock; /* if writing end-of-archive, aid compression (good hygiene too) */ if (curblk < endblk) memset(curblk, 0, (char *)endblk - (char *)curblk); if (write(ar, tpblk, bytes) != bytes) sysfatal("error writing archive: %r"); } static void putblk(int ar) { if (curblk >= endblk) putlastblk(ar); } static void putbackblk(int ar) { curblk--; USED(ar); } static void putreadblks(int ar, int blks) { curblk += blks - 1; USED(ar); } static void putblkmany(int ar, int blks) { curblk += blks - 1; putblk(ar); } /* * common routines */ /* * modifies hp->chksum but restores it; important for the last block of the * old archive when updating with `tar rf archive' */ static long chksum(Hdr *hp) { int n = Tblock; long i = 0; uchar *cp = (uchar*)hp; char oldsum[sizeof hp->chksum]; memmove(oldsum, hp->chksum, sizeof oldsum); memset(hp->chksum, ' ', sizeof hp->chksum); while (n-- > 0) i += *cp++; memmove(hp->chksum, oldsum, sizeof oldsum); return i; } static int isustar(Hdr *hp) { return strcmp(hp->magic, "ustar") == 0; } /* * s is at most n bytes long, but need not be NUL-terminated. * if shorter than n bytes, all bytes after the first NUL must also * be NUL. */ static int sstrnlen(char *s, int n) { return s[n - 1] != '\0'? n: strlen(s); } /* set fullname from header */ static char * name(Hdr *hp) { int pfxlen, namlen; static char fullnamebuf[2 + Maxname + 1]; /* 2 at beginning for ./ on relative names */ char *fullname; fullname = fullnamebuf+2; namlen = sstrnlen(hp->name, sizeof hp->name); if (hp->prefix[0] == '\0' || !isustar(hp)) { /* old-style name? */ memmove(fullname, hp->name, namlen); fullname[namlen] = '\0'; return fullname; } /* name is in two pieces */ pfxlen = sstrnlen(hp->prefix, sizeof hp->prefix); memmove(fullname, hp->prefix, pfxlen); fullname[pfxlen] = '/'; memmove(fullname + pfxlen + 1, hp->name, namlen); fullname[pfxlen + 1 + namlen] = '\0'; return fullname; } static int isdir(Hdr *hp) { /* the mode test is ugly but sometimes necessary */ return hp->linkflag == LF_DIR || strrchr(name(hp), '\0')[-1] == '/' || (strtoul(hp->mode, nil, 8)&0170000) == 040000; } static int eotar(Hdr *hp) { return name(hp)[0] == '\0'; } /* static uvlong getbe(uchar *src, int size) { uvlong vl = 0; while (size-- > 0) { vl <<= 8; vl |= *src++; } return vl; } */ static void putbe(uchar *dest, uvlong vl, int size) { for (dest += size; size-- > 0; vl >>= 8) *--dest = vl; } /* * return the nominal size from the header block, which is not always the * size in the archive (the archive size may be zero for some file types * regardless of the nominal size). * * gnu and freebsd tars are now recording vlongs as big-endian binary * with a flag in byte 0 to indicate this, which permits file sizes up to * 2^64-1 (actually 2^80-1 but our file sizes are vlongs) rather than 2^33-1. */ static Off hdrsize(Hdr *hp) { uchar *p; if((uchar)hp->size[0] == Binnegsz) { fprint(2, "%s: %s: negative length, which is insane\n", argv0, name(hp)); return 0; } else if((uchar)hp->size[0] == Binsize) { p = (uchar *)hp->size + sizeof hp->size - 1 - sizeof(vlong); /* -1 for terminating space */