From 7b0ad9c638176dc364dabb65b363536055a0ea9c Mon Sep 17 00:00:00 2001 From: Tianjie Xu Date: Fri, 5 Aug 2016 18:00:04 -0700 Subject: Switch recovery to libbase logging Clean up the recovery image and switch to libbase logging. Bug: 28191554 Change-Id: Icd999c3cc832f0639f204b5c36cea8afe303ad35 Merged-In: Icd999c3cc832f0639f204b5c36cea8afe303ad35 --- minzip/Zip.cpp | 1022 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 1022 insertions(+) create mode 100644 minzip/Zip.cpp (limited to 'minzip/Zip.cpp') diff --git a/minzip/Zip.cpp b/minzip/Zip.cpp new file mode 100644 index 000000000..b887b8466 --- /dev/null +++ b/minzip/Zip.cpp @@ -0,0 +1,1022 @@ +/* + * Copyright 2006 The Android Open Source Project + * + * Simple Zip file support. + */ +#include "safe_iop.h" +#include "zlib.h" + +#include +#include +#include +#include // for uintptr_t +#include +#include // for S_ISLNK() +#include + +#include + +#include +#include +#include +#include +#include + +#include "Zip.h" +#include "Bits.h" +#include "DirUtil.h" + +#define SORT_ENTRIES 1 + +/* + * Offset and length constants (java.util.zip naming convention). + */ +enum { + CENSIG = 0x02014b50, // PK12 + CENHDR = 46, + + CENVEM = 4, + CENVER = 6, + CENFLG = 8, + CENHOW = 10, + CENTIM = 12, + CENCRC = 16, + CENSIZ = 20, + CENLEN = 24, + CENNAM = 28, + CENEXT = 30, + CENCOM = 32, + CENDSK = 34, + CENATT = 36, + CENATX = 38, + CENOFF = 42, + + ENDSIG = 0x06054b50, // PK56 + ENDHDR = 22, + + ENDSUB = 8, + ENDTOT = 10, + ENDSIZ = 12, + ENDOFF = 16, + ENDCOM = 20, + + EXTSIG = 0x08074b50, // PK78 + EXTHDR = 16, + + EXTCRC = 4, + EXTSIZ = 8, + EXTLEN = 12, + + LOCSIG = 0x04034b50, // PK34 + LOCHDR = 30, + + LOCVER = 4, + LOCFLG = 6, + LOCHOW = 8, + LOCTIM = 10, + LOCCRC = 14, + LOCSIZ = 18, + LOCLEN = 22, + LOCNAM = 26, + LOCEXT = 28, + + STORED = 0, + DEFLATED = 8, + + CENVEM_UNIX = 3 << 8, // the high byte of CENVEM +}; + + +/* + * For debugging, dump the contents of a ZipEntry. + */ +#if 0 +static void dumpEntry(const ZipEntry* pEntry) +{ + LOGI(" %p '%.*s'\n", pEntry->fileName,pEntry->fileNameLen,pEntry->fileName); + LOGI(" off=%u comp=%u uncomp=%u how=%d\n", pEntry->offset, + pEntry->compLen, pEntry->uncompLen, pEntry->compression); +} +#endif + +/* + * (This is a mzHashTableLookup callback.) + * + * Compare two ZipEntry structs, by name. + */ +static int hashcmpZipEntry(const void* ventry1, const void* ventry2) +{ + const ZipEntry* entry1 = (const ZipEntry*) ventry1; + const ZipEntry* entry2 = (const ZipEntry*) ventry2; + + if (entry1->fileNameLen != entry2->fileNameLen) + return entry1->fileNameLen - entry2->fileNameLen; + return memcmp(entry1->fileName, entry2->fileName, entry1->fileNameLen); +} + +/* + * (This is a mzHashTableLookup callback.) + * + * find a ZipEntry struct by name. + */ +static int hashcmpZipName(const void* ventry, const void* vname) +{ + const ZipEntry* entry = (const ZipEntry*) ventry; + const char* name = (const char*) vname; + unsigned int nameLen = strlen(name); + + if (entry->fileNameLen != nameLen) + return entry->fileNameLen - nameLen; + return memcmp(entry->fileName, name, nameLen); +} + +/* + * Compute the hash code for a ZipEntry filename. + * + * Not expected to be compatible with any other hash function, so we init + * to 2 to ensure it doesn't happen to match. + */ +static unsigned int computeHash(const char* name, int nameLen) +{ + unsigned int hash = 2; + + while (nameLen--) + hash = hash * 31 + *name++; + + return hash; +} + +static void addEntryToHashTable(HashTable* pHash, ZipEntry* pEntry) +{ + unsigned int itemHash = computeHash(pEntry->fileName, pEntry->fileNameLen); + const ZipEntry* found; + + found = (const ZipEntry*)mzHashTableLookup(pHash, + itemHash, pEntry, hashcmpZipEntry, true); + if (found != pEntry) { + LOG(WARNING) << "WARNING: duplicate entry '" << std::string(found->fileName, + found->fileNameLen) << "' in Zip"; + + /* keep going */ + } +} + +static int validFilename(const char *fileName, unsigned int fileNameLen) +{ + // Forbid super long filenames. + if (fileNameLen >= PATH_MAX) { + LOG(WARNING) << "Filename too long (" << fileNameLen << " chatacters)"; + return 0; + } + + // Require all characters to be printable ASCII (no NUL, no UTF-8, etc). + unsigned int i; + for (i = 0; i < fileNameLen; ++i) { + if (fileName[i] < 32 || fileName[i] >= 127) { + LOG(WARNING) << android::base::StringPrintf( + "Filename contains invalid character '\%02x'\n", fileName[i]); + return 0; + } + } + + return 1; +} + +/* + * Parse the contents of a Zip archive. After confirming that the file + * is in fact a Zip, we scan out the contents of the central directory and + * store it in a hash table. + * + * Returns "true" on success. + */ +static bool parseZipArchive(ZipArchive* pArchive) +{ + bool result = false; + const unsigned char* ptr; + unsigned int i, numEntries, cdOffset; + unsigned int val; + + /* + * The first 4 bytes of the file will either be the local header + * signature for the first file (LOCSIG) or, if the archive doesn't + * have any files in it, the end-of-central-directory signature (ENDSIG). + */ + val = get4LE(pArchive->addr); + if (val == ENDSIG) { + LOG(WARNING) << "Found Zip archive, but it looks empty"; + goto bail; + } else if (val != LOCSIG) { + LOG(WARNING) << android::base::StringPrintf("Not a Zip archive (found 0x%08x)\n", val); + goto bail; + } + + /* + * Find the EOCD. We'll find it immediately unless they have a file + * comment. + */ + ptr = pArchive->addr + pArchive->length - ENDHDR; + + while (ptr >= (const unsigned char*) pArchive->addr) { + if (*ptr == (ENDSIG & 0xff) && get4LE(ptr) == ENDSIG) + break; + ptr--; + } + if (ptr < (const unsigned char*) pArchive->addr) { + LOG(WARNING) << "Could not find end-of-central-directory in Zip"; + goto bail; + } + + /* + * There are two interesting items in the EOCD block: the number of + * entries in the file, and the file offset of the start of the + * central directory. + */ + numEntries = get2LE(ptr + ENDSUB); + cdOffset = get4LE(ptr + ENDOFF); + + LOG(VERBOSE) << "numEntries=" << numEntries << " cdOffset=" << cdOffset; + if (numEntries == 0 || cdOffset >= pArchive->length) { + LOG(WARNING) << "Invalid entries=" << numEntries << " offset=" << cdOffset + << " (len=" << pArchive->length << ")"; + goto bail; + } + + /* + * Create data structures to hold entries. + */ + pArchive->numEntries = numEntries; + pArchive->pEntries = (ZipEntry*) calloc(numEntries, sizeof(ZipEntry)); + pArchive->pHash = mzHashTableCreate(mzHashSize(numEntries), NULL); + if (pArchive->pEntries == NULL || pArchive->pHash == NULL) + goto bail; + + ptr = pArchive->addr + cdOffset; + for (i = 0; i < numEntries; i++) { + ZipEntry* pEntry; + unsigned int fileNameLen, extraLen, commentLen, localHdrOffset; + const unsigned char* localHdr; + const char *fileName; + + if (ptr + CENHDR > (const unsigned char*)pArchive->addr + pArchive->length) { + LOG(WARNING) << "Ran off the end (at " << i << ")"; + goto bail; + } + if (get4LE(ptr) != CENSIG) { + LOG(WARNING) << "Missed a central dir sig (at " << i << ")"; + goto bail; + } + + localHdrOffset = get4LE(ptr + CENOFF); + fileNameLen = get2LE(ptr + CENNAM); + extraLen = get2LE(ptr + CENEXT); + commentLen = get2LE(ptr + CENCOM); + fileName = (const char*)ptr + CENHDR; + if (fileName + fileNameLen > (const char*)pArchive->addr + pArchive->length) { + LOG(WARNING) << "Filename ran off the end (at " << i << ")"; + goto bail; + } + if (!validFilename(fileName, fileNameLen)) { + LOG(WARNING) << "Invalid filename (at " << i << ")"; + goto bail; + } + +#if SORT_ENTRIES + /* Figure out where this entry should go (binary search). + */ + if (i > 0) { + int low, high; + + low = 0; + high = i - 1; + while (low <= high) { + int mid; + int diff; + int diffLen; + + mid = low + ((high - low) / 2); // avoid overflow + + if (pArchive->pEntries[mid].fileNameLen < fileNameLen) { + diffLen = pArchive->pEntries[mid].fileNameLen; + } else { + diffLen = fileNameLen; + } + diff = strncmp(pArchive->pEntries[mid].fileName, fileName, + diffLen); + if (diff == 0) { + diff = pArchive->pEntries[mid].fileNameLen - fileNameLen; + } + if (diff < 0) { + low = mid + 1; + } else if (diff > 0) { + high = mid - 1; + } else { + high = mid; + break; + } + } + + unsigned int target = high + 1; + assert(target <= i); + if (target != i) { + /* It belongs somewhere other than at the end of + * the list. Make some room at [target]. + */ + memmove(pArchive->pEntries + target + 1, + pArchive->pEntries + target, + (i - target) * sizeof(ZipEntry)); + } + pEntry = &pArchive->pEntries[target]; + } else { + pEntry = &pArchive->pEntries[0]; + } +#else + pEntry = &pArchive->pEntries[i]; +#endif + pEntry->fileNameLen = fileNameLen; + pEntry->fileName = fileName; + + pEntry->compLen = get4LE(ptr + CENSIZ); + pEntry->uncompLen = get4LE(ptr + CENLEN); + pEntry->compression = get2LE(ptr + CENHOW); + pEntry->modTime = get4LE(ptr + CENTIM); + pEntry->crc32 = get4LE(ptr + CENCRC); + + /* These two are necessary for finding the mode of the file. + */ + pEntry->versionMadeBy = get2LE(ptr + CENVEM); + if ((pEntry->versionMadeBy & 0xff00) != 0 && + (pEntry->versionMadeBy & 0xff00) != CENVEM_UNIX) + { + LOG(WARNING) << android::base::StringPrintf( + "Incompatible \"version made by\": 0x%02x (at %d)\n", + pEntry->versionMadeBy >> 8, i); + goto bail; + } + pEntry->externalFileAttributes = get4LE(ptr + CENATX); + + // Perform pArchive->addr + localHdrOffset, ensuring that it won't + // overflow. This is needed because localHdrOffset is untrusted. + if (!safe_add((uintptr_t *)&localHdr, (uintptr_t)pArchive->addr, + (uintptr_t)localHdrOffset)) { + LOG(WARNING) << "Integer overflow adding in parseZipArchive"; + goto bail; + } + if ((uintptr_t)localHdr + LOCHDR > + (uintptr_t)pArchive->addr + pArchive->length) { + LOG(WARNING) << "Bad offset to local header: " << localHdrOffset + << " (at " << i << ")"; + goto bail; + } + if (get4LE(localHdr) != LOCSIG) { + LOG(WARNING) << "Missed a local header sig (at " << i << ")"; + goto bail; + } + pEntry->offset = localHdrOffset + LOCHDR + + get2LE(localHdr + LOCNAM) + get2LE(localHdr + LOCEXT); + if (!safe_add(NULL, pEntry->offset, pEntry->compLen)) { + LOG(WARNING) << "Integer overflow adding in parseZipArchive"; + goto bail; + } + if ((size_t)pEntry->offset + pEntry->compLen > pArchive->length) { + LOG(WARNING) << "Data ran off the end (at " << i << ")"; + goto bail; + } + +#if !SORT_ENTRIES + /* Add to hash table; no need to lock here. + * Can't do this now if we're sorting, because entries + * will move around. + */ + addEntryToHashTable(pArchive->pHash, pEntry); +#endif + + //dumpEntry(pEntry); + ptr += CENHDR + fileNameLen + extraLen + commentLen; + } + +#if SORT_ENTRIES + /* If we're sorting, we have to wait until all entries + * are in their final places, otherwise the pointers will + * probably point to the wrong things. + */ + for (i = 0; i < numEntries; i++) { + /* Add to hash table; no need to lock here. + */ + addEntryToHashTable(pArchive->pHash, &pArchive->pEntries[i]); + } +#endif + + result = true; + +bail: + if (!result) { + mzHashTableFree(pArchive->pHash); + pArchive->pHash = NULL; + } + return result; +} + +/* + * Open a Zip archive and scan out the contents. + * + * The easiest way to do this is to mmap() the whole thing and do the + * traditional backward scan for central directory. Since the EOCD is + * a relatively small bit at the end, we should end up only touching a + * small set of pages. + * + * This will be called on non-Zip files, especially during startup, so + * we don't want to be too noisy about failures. (Do we want a "quiet" + * flag?) + * + * On success, we fill out the contents of "pArchive". + */ +int mzOpenZipArchive(unsigned char* addr, size_t length, ZipArchive* pArchive) +{ + int err; + + if (length < ENDHDR) { + err = -1; + LOG(WARNING) << "Archive " << pArchive << " is too small to be zip (" + << length << ")"; + goto bail; + } + + pArchive->addr = addr; + pArchive->length = length; + + if (!parseZipArchive(pArchive)) { + err = -1; + LOG(WARNING) << "Parsing archive " << pArchive << " failed"; + goto bail; + } + + err = 0; + +bail: + if (err != 0) + mzCloseZipArchive(pArchive); + return err; +} + +/* + * Close a ZipArchive, closing the file and freeing the contents. + * + * NOTE: the ZipArchive may not have been fully created. + */ +void mzCloseZipArchive(ZipArchive* pArchive) +{ + LOG(VERBOSE) << "Closing archive " << pArchive; + + free(pArchive->pEntries); + + mzHashTableFree(pArchive->pHash); + + pArchive->pHash = NULL; + pArchive->pEntries = NULL; +} + +/* + * Find a matching entry. + * + * Returns NULL if no matching entry found. + */ +const ZipEntry* mzFindZipEntry(const ZipArchive* pArchive, + const char* entryName) +{ + unsigned int itemHash = computeHash(entryName, strlen(entryName)); + + return (const ZipEntry*)mzHashTableLookup(pArchive->pHash, + itemHash, (char*) entryName, hashcmpZipName, false); +} + +/* + * Return true if the entry is a symbolic link. + */ +static bool mzIsZipEntrySymlink(const ZipEntry* pEntry) +{ + if ((pEntry->versionMadeBy & 0xff00) == CENVEM_UNIX) { + return S_ISLNK(pEntry->externalFileAttributes >> 16); + } + return false; +} + +/* Call processFunction on the uncompressed data of a STORED entry. + */ +static bool processStoredEntry(const ZipArchive *pArchive, + const ZipEntry *pEntry, ProcessZipEntryContentsFunction processFunction, + void *cookie) +{ + return processFunction(pArchive->addr + pEntry->offset, pEntry->uncompLen, cookie); +} + +static bool processDeflatedEntry(const ZipArchive *pArchive, + const ZipEntry *pEntry, ProcessZipEntryContentsFunction processFunction, + void *cookie) +{ + bool success = false; + unsigned long totalOut = 0; + unsigned char procBuf[32 * 1024]; + z_stream zstream; + int zerr; + + /* + * Initialize the zlib stream. + */ + memset(&zstream, 0, sizeof(zstream)); + zstream.zalloc = Z_NULL; + zstream.zfree = Z_NULL; + zstream.opaque = Z_NULL; + zstream.next_in = pArchive->addr + pEntry->offset; + zstream.avail_in = pEntry->compLen; + zstream.next_out = (Bytef*) procBuf; + zstream.avail_out = sizeof(procBuf); + zstream.data_type = Z_UNKNOWN; + + /* + * Use the undocumented "negative window bits" feature to tell zlib + * that there's no zlib header waiting for it. + */ + zerr = inflateInit2(&zstream, -MAX_WBITS); + if (zerr != Z_OK) { + if (zerr == Z_VERSION_ERROR) { + LOG(ERROR) << "Installed zlib is not compatible with linked version (" + << ZLIB_VERSION << ")"; + } else { + LOG(ERROR) << "Call to inflateInit2 failed (zerr=" << zerr << ")"; + } + goto bail; + } + + /* + * Loop while we have data. + */ + do { + /* uncompress the data */ + zerr = inflate(&zstream, Z_NO_FLUSH); + if (zerr != Z_OK && zerr != Z_STREAM_END) { + LOG(WARNING) << "zlib inflate call failed (zerr=" << zerr << ")"; + goto z_bail; + } + + /* write when we're full or when we're done */ + if (zstream.avail_out == 0 || + (zerr == Z_STREAM_END && zstream.avail_out != sizeof(procBuf))) + { + long procSize = zstream.next_out - procBuf; + LOG(VERBOSE) << "+++ processing " << procSize << " bytes"; + bool ret = processFunction(procBuf, procSize, cookie); + if (!ret) { + LOG(WARNING) << "Process function elected to fail (in inflate)"; + goto z_bail; + } + + zstream.next_out = procBuf; + zstream.avail_out = sizeof(procBuf); + } + } while (zerr == Z_OK); + + assert(zerr == Z_STREAM_END); /* other errors should've been caught */ + + // success! + totalOut = zstream.total_out; + success = true; + +z_bail: + inflateEnd(&zstream); /* free up any allocated structures */ + +bail: + if (totalOut != pEntry->uncompLen) { + if (success) { // error already shown? + LOG(WARNING) << "Size mismatch on inflated file (" << totalOut << " vs " + << pEntry->uncompLen << ")"; + } + return false; + } + return true; +} + +/* + * Stream the uncompressed data through the supplied function, + * passing cookie to it each time it gets called. processFunction + * may be called more than once. + * + * If processFunction returns false, the operation is abandoned and + * mzProcessZipEntryContents() immediately returns false. + * + * This is useful for calculating the hash of an entry's uncompressed contents. + */ +bool mzProcessZipEntryContents(const ZipArchive *pArchive, + const ZipEntry *pEntry, ProcessZipEntryContentsFunction processFunction, + void *cookie) +{ + bool ret = false; + + switch (pEntry->compression) { + case STORED: + ret = processStoredEntry(pArchive, pEntry, processFunction, cookie); + break; + case DEFLATED: + ret = processDeflatedEntry(pArchive, pEntry, processFunction, cookie); + break; + default: + LOG(ERROR) << "Unsupported compression type " << pEntry->compression + << " for entry '" << pEntry->fileName << "'"; + break; + } + + return ret; +} + +typedef struct { + char *buf; + int bufLen; +} CopyProcessArgs; + +static bool copyProcessFunction(const unsigned char *data, int dataLen, + void *cookie) +{ + CopyProcessArgs *args = (CopyProcessArgs *)cookie; + if (dataLen <= args->bufLen) { + memcpy(args->buf, data, dataLen); + args->buf += dataLen; + args->bufLen -= dataLen; + return true; + } + return false; +} + +/* + * Read an entry into a buffer allocated by the caller. + */ +bool mzReadZipEntry(const ZipArchive* pArchive, const ZipEntry* pEntry, + char *buf, int bufLen) +{ + CopyProcessArgs args; + bool ret; + + args.buf = buf; + args.bufLen = bufLen; + ret = mzProcessZipEntryContents(pArchive, pEntry, copyProcessFunction, + (void *)&args); + if (!ret) { + LOG(ERROR) << "Can't extract entry to buffer"; + return false; + } + return true; +} + +static bool writeProcessFunction(const unsigned char *data, int dataLen, + void *cookie) +{ + int fd = (int)(intptr_t)cookie; + if (dataLen == 0) { + return true; + } + ssize_t soFar = 0; + while (true) { + ssize_t n = TEMP_FAILURE_RETRY(write(fd, data+soFar, dataLen-soFar)); + if (n <= 0) { + PLOG(ERROR) << "Error writing " << dataLen-soFar << " bytes from zip file from " + << data+soFar; + return false; + } else if (n > 0) { + soFar += n; + if (soFar == dataLen) return true; + if (soFar > dataLen) { + LOG(ERROR) << "write overrun? (" << soFar << " bytes instead of " + << dataLen << ")"; + return false; + } + } + } +} + +/* + * Uncompress "pEntry" in "pArchive" to "fd" at the current offset. + */ +bool mzExtractZipEntryToFile(const ZipArchive *pArchive, + const ZipEntry *pEntry, int fd) +{ + bool ret = mzProcessZipEntryContents(pArchive, pEntry, writeProcessFunction, + (void*)(intptr_t)fd); + if (!ret) { + LOG(ERROR) << "Can't extract entry to file."; + return false; + } + return true; +} + +typedef struct { + unsigned char* buffer; + long len; +} BufferExtractCookie; + +static bool bufferProcessFunction(const unsigned char *data, int dataLen, + void *cookie) { + BufferExtractCookie *bec = (BufferExtractCookie*)cookie; + + memmove(bec->buffer, data, dataLen); + bec->buffer += dataLen; + bec->len -= dataLen; + + return true; +} + +/* + * Uncompress "pEntry" in "pArchive" to buffer, which must be large + * enough to hold mzGetZipEntryUncomplen(pEntry) bytes. + */ +bool mzExtractZipEntryToBuffer(const ZipArchive *pArchive, + const ZipEntry *pEntry, unsigned char *buffer) +{ + BufferExtractCookie bec; + bec.buffer = buffer; + bec.len = mzGetZipEntryUncompLen(pEntry); + + bool ret = mzProcessZipEntryContents(pArchive, pEntry, + bufferProcessFunction, (void*)&bec); + if (!ret || bec.len != 0) { + LOG(ERROR) << "Can't extract entry to memory buffer."; + return false; + } + return true; +} + + +/* Helper state to make path translation easier and less malloc-happy. + */ +typedef struct { + const char *targetDir; + const char *zipDir; + char *buf; + int targetDirLen; + int zipDirLen; + int bufLen; +} MzPathHelper; + +/* Given the values of targetDir and zipDir in the helper, + * return the target filename of the provided entry. + * The helper must be initialized first. + */ +static const char *targetEntryPath(MzPathHelper *helper, ZipEntry *pEntry) +{ + int needLen; + bool firstTime = (helper->buf == NULL); + + /* target file <-- targetDir + / + entry[zipDirLen:] + */ + needLen = helper->targetDirLen + 1 + + pEntry->fileNameLen - helper->zipDirLen + 1; + if (firstTime || needLen > helper->bufLen) { + char *newBuf; + + needLen *= 2; + newBuf = (char *)realloc(helper->buf, needLen); + if (newBuf == NULL) { + return NULL; + } + helper->buf = newBuf; + helper->bufLen = needLen; + } + + /* Every path will start with the target path and a slash. + */ + if (firstTime) { + char *p = helper->buf; + memcpy(p, helper->targetDir, helper->targetDirLen); + p += helper->targetDirLen; + if (p == helper->buf || p[-1] != '/') { + helper->targetDirLen += 1; + *p++ = '/'; + } + } + + /* Replace the custom part of the path with the appropriate + * part of the entry's path. + */ + char *epath = helper->buf + helper->targetDirLen; + memcpy(epath, pEntry->fileName + helper->zipDirLen, + pEntry->fileNameLen - helper->zipDirLen); + epath += pEntry->fileNameLen - helper->zipDirLen; + *epath = '\0'; + + return helper->buf; +} + +/* + * Inflate all entries under zipDir to the directory specified by + * targetDir, which must exist and be a writable directory. + * + * The immediate children of zipDir will become the immediate + * children of targetDir; e.g., if the archive contains the entries + * + * a/b/c/one + * a/b/c/two + * a/b/c/d/three + * + * and mzExtractRecursive(a, "a/b/c", "/tmp") is called, the resulting + * files will be + * + * /tmp/one + * /tmp/two + * /tmp/d/three + * + * Returns true on success, false on failure. + */ +bool mzExtractRecursive(const ZipArchive *pArchive, + const char *zipDir, const char *targetDir, + const struct utimbuf *timestamp, + void (*callback)(const char *fn, void *), void *cookie, + struct selabel_handle *sehnd) +{ + if (zipDir[0] == '/') { + LOG(ERROR) << "mzExtractRecursive(): zipDir must be a relative path."; + return false; + } + if (targetDir[0] != '/') { + LOG(ERROR) << "mzExtractRecursive(): targetDir must be an absolute path.\n"; + return false; + } + + unsigned int zipDirLen; + char *zpath; + + zipDirLen = strlen(zipDir); + zpath = (char *)malloc(zipDirLen + 2); + if (zpath == NULL) { + LOG(ERROR) << "Can't allocate " << (zipDirLen + 2) << " bytes for zip path"; + return false; + } + /* If zipDir is empty, we'll extract the entire zip file. + * Otherwise, canonicalize the path. + */ + if (zipDirLen > 0) { + /* Make sure there's (hopefully, exactly one) slash at the + * end of the path. This way we don't need to worry about + * accidentally extracting "one/twothree" when a path like + * "one/two" is specified. + */ + memcpy(zpath, zipDir, zipDirLen); + if (zpath[zipDirLen-1] != '/') { + zpath[zipDirLen++] = '/'; + } + } + zpath[zipDirLen] = '\0'; + + /* Set up the helper structure that we'll use to assemble paths. + */ + MzPathHelper helper; + helper.targetDir = targetDir; + helper.targetDirLen = strlen(helper.targetDir); + helper.zipDir = zpath; + helper.zipDirLen = strlen(helper.zipDir); + helper.buf = NULL; + helper.bufLen = 0; + + /* Walk through the entries and extract anything whose path begins + * with zpath. + //TODO: since the entries are sorted, binary search for the first match + // and stop after the first non-match. + */ + unsigned int i; + bool seenMatch = false; + int ok = true; + int extractCount = 0; + for (i = 0; i < pArchive->numEntries; i++) { + ZipEntry *pEntry = pArchive->pEntries + i; + if (pEntry->fileNameLen < zipDirLen) { + //TODO: look out for a single empty directory entry that matches zpath, but + // missing the trailing slash. Most zip files seem to include + // the trailing slash, but I think it's legal to leave it off. + // e.g., zpath "a/b/", entry "a/b", with no children of the entry. + /* No chance of matching. + */ +#if SORT_ENTRIES + if (seenMatch) { + /* Since the entries are sorted, we can give up + * on the first mismatch after the first match. + */ + break; + } +#endif + continue; + } + /* If zpath is empty, this strncmp() will match everything, + * which is what we want. + */ + if (strncmp(pEntry->fileName, zpath, zipDirLen) != 0) { +#if SORT_ENTRIES + if (seenMatch) { + /* Since the entries are sorted, we can give up + * on the first mismatch after the first match. + */ + break; + } +#endif + continue; + } + /* This entry begins with zipDir, so we'll extract it. + */ + seenMatch = true; + + /* Find the target location of the entry. + */ + const char *targetFile = targetEntryPath(&helper, pEntry); + if (targetFile == NULL) { + LOG(ERROR) << "Can't assemble target path for \"" << std::string(pEntry->fileName, + pEntry->fileNameLen) << "\""; + ok = false; + break; + } + +#define UNZIP_DIRMODE 0755 +#define UNZIP_FILEMODE 0644 + /* + * Create the file or directory. We ignore directory entries + * because we recursively create paths to each file entry we encounter + * in the zip archive anyway. + * + * NOTE: A "directory entry" in a zip archive is just a zero length + * entry that ends in a "/". They're not mandatory and many tools get + * rid of them. We need to process them only if we want to preserve + * empty directories from the archive. + */ + if (pEntry->fileName[pEntry->fileNameLen-1] != '/') { + /* This is not a directory. First, make sure that + * the containing directory exists. + */ + int ret = dirCreateHierarchy( + targetFile, UNZIP_DIRMODE, timestamp, true, sehnd); + if (ret != 0) { + PLOG(ERROR) << "Can't create containing directory for \"" << targetFile << "\""; + ok = false; + break; + } + + /* + * The entry is a regular file or a symlink. Open the target for writing. + * + * TODO: This behavior for symlinks seems rather bizarre. For a + * symlink foo/bar/baz -> foo/tar/taz, we will create a file called + * "foo/bar/baz" whose contents are the literal "foo/tar/taz". We + * warn about this for now and preserve older behavior. + */ + if (mzIsZipEntrySymlink(pEntry)) { + LOG(ERROR) << "Symlink entry \"" << std::string(pEntry->fileName, + pEntry->fileNameLen) << "\" will be output as a regular file."; + } + + char *secontext = NULL; + + if (sehnd) { + selabel_lookup(sehnd, &secontext, targetFile, UNZIP_FILEMODE); + setfscreatecon(secontext); + } + + int fd = open(targetFile, O_CREAT|O_WRONLY|O_TRUNC, + UNZIP_FILEMODE); + + if (secontext) { + freecon(secontext); + setfscreatecon(NULL); + } + + if (fd < 0) { + PLOG(ERROR) << "Can't create target file \"" << targetFile << "\""; + ok = false; + break; + } + + bool ok = mzExtractZipEntryToFile(pArchive, pEntry, fd); + if (ok) { + ok = (fsync(fd) == 0); + } + if (close(fd) != 0) { + ok = false; + } + if (!ok) { + LOG(ERROR) << "Error extracting \"" << targetFile << "\""; + ok = false; + break; + } + + if (timestamp != NULL && utime(targetFile, timestamp)) { + LOG(ERROR) << "Error touching \"" << targetFile << "\""; + ok = false; + break; + } + + LOG(VERBOSE) <<"Extracted file \"" << targetFile << "\""; + ++extractCount; + } + + if (callback != NULL) callback(targetFile, cookie); + } + + LOG(VERBOSE) << "Extracted " << extractCount << " file(s)"; + + free(helper.buf); + free(zpath); + + return ok; +} -- cgit v1.2.3