blob: 7da9adf5f1014d4ebf92ce5729cd13b714975ff9 [file] [log] [blame]
/*
* Copyright (C) 2014 The Android Open Source Project
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
#include <ctype.h>
#include <errno.h>
#include <dirent.h>
#include <fcntl.h>
#include <inttypes.h>
#include <linux/fs.h>
#include <pthread.h>
#include <stdarg.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <sys/ioctl.h>
#include <time.h>
#include <unistd.h>
#include <memory>
#include <string>
#include <base/strings.h>
#include "applypatch/applypatch.h"
#include "edify/expr.h"
#include "mincrypt/sha.h"
#include "minzip/Hash.h"
#include "updater.h"
#include "print_sha1.h"
#define BLOCKSIZE 4096
// Set this to 0 to interpret 'erase' transfers to mean do a
// BLKDISCARD ioctl (the normal behavior). Set to 1 to interpret
// erase to mean fill the region with zeroes.
#define DEBUG_ERASE 0
#define STASH_DIRECTORY_BASE "/cache/recovery"
#define STASH_DIRECTORY_MODE 0700
#define STASH_FILE_MODE 0600
typedef struct {
size_t count;
size_t size;
size_t pos[0]; // Actual limit is INT_MAX.
} RangeSet;
#define RANGESET_MAX_POINTS \
((int)((INT_MAX / sizeof(int)) - sizeof(RangeSet)))
static RangeSet* parse_range(char* text) {
char* save;
char* token;
int num;
long int val;
RangeSet* out = NULL;
size_t bufsize;
if (!text) {
goto err;
}
token = strtok_r(text, ",", &save);
if (!token) {
goto err;
}
errno = 0;
val = strtol(token, NULL, 0);
if (errno != 0 || val < 2 || val > RANGESET_MAX_POINTS) {
goto err;
} else if (val % 2) {
goto err; // must be even
}
num = (int) val;
bufsize = sizeof(RangeSet) + num * sizeof(size_t);
out = reinterpret_cast<RangeSet*>(malloc(bufsize));
if (!out) {
fprintf(stderr, "failed to allocate range of %zu bytes\n", bufsize);
goto err;
}
out->count = num / 2;
out->size = 0;
for (int i = 0; i < num; i += 2) {
token = strtok_r(NULL, ",", &save);
if (!token) {
goto err;
}
errno = 0;
val = strtol(token, NULL, 0);
if (errno != 0 || val < 0 || val > INT_MAX) {
goto err;
}
out->pos[i] = static_cast<size_t>(val);
token = strtok_r(NULL, ",", &save);
if (!token) {
goto err;
}
errno = 0;
val = strtol(token, NULL, 0);
if (errno != 0 || val < 0 || val > INT_MAX) {
goto err;
}
out->pos[i+1] = static_cast<size_t>(val);
if (out->pos[i] >= out->pos[i+1]) {
goto err; // empty or negative range
}
size_t rs = out->pos[i+1] - out->pos[i];
if (out->size > SIZE_MAX - rs) {
goto err; // overflow
}
out->size += rs;
}
if (out->size == 0) {
goto err;
}
return out;
err:
fprintf(stderr, "failed to parse range '%s'\n", text ? text : "NULL");
exit(1);
}
static bool range_overlaps(const RangeSet& r1, const RangeSet& r2) {
for (size_t i = 0; i < r1.count; ++i) {
size_t r1_0 = r1.pos[i * 2];
size_t r1_1 = r1.pos[i * 2 + 1];
for (size_t j = 0; j < r2.count; ++j) {
size_t r2_0 = r2.pos[j * 2];
size_t r2_1 = r2.pos[j * 2 + 1];
if (!(r2_0 >= r1_1 || r1_0 >= r2_1)) {
return true;
}
}
}
return false;
}
static int read_all(int fd, uint8_t* data, size_t size) {
size_t so_far = 0;
while (so_far < size) {
ssize_t r = TEMP_FAILURE_RETRY(read(fd, data+so_far, size-so_far));
if (r == -1) {
fprintf(stderr, "read failed: %s\n", strerror(errno));
return -1;
}
so_far += r;
}
return 0;
}
static int write_all(int fd, const uint8_t* data, size_t size) {
size_t written = 0;
while (written < size) {
ssize_t w = TEMP_FAILURE_RETRY(write(fd, data+written, size-written));
if (w == -1) {
fprintf(stderr, "write failed: %s\n", strerror(errno));
return -1;
}
written += w;
}
return 0;
}
static bool check_lseek(int fd, off64_t offset, int whence) {
off64_t rc = TEMP_FAILURE_RETRY(lseek64(fd, offset, whence));
if (rc == -1) {
fprintf(stderr, "lseek64 failed: %s\n", strerror(errno));
return false;
}
return true;
}
static void allocate(size_t size, uint8_t** buffer, size_t* buffer_alloc) {
// if the buffer's big enough, reuse it.
if (size <= *buffer_alloc) return;
free(*buffer);
*buffer = (uint8_t*) malloc(size);
if (*buffer == NULL) {
fprintf(stderr, "failed to allocate %zu bytes\n", size);
exit(1);
}
*buffer_alloc = size;
}
typedef struct {
int fd;
RangeSet* tgt;
size_t p_block;
size_t p_remain;
} RangeSinkState;
static ssize_t RangeSinkWrite(const uint8_t* data, ssize_t size, void* token) {
RangeSinkState* rss = (RangeSinkState*) token;
if (rss->p_remain <= 0) {
fprintf(stderr, "range sink write overrun");
return 0;
}
ssize_t written = 0;
while (size > 0) {
size_t write_now = size;
if (rss->p_remain < write_now) {
write_now = rss->p_remain;
}
if (write_all(rss->fd, data, write_now) == -1) {
break;
}
data += write_now;
size -= write_now;
rss->p_remain -= write_now;
written += write_now;
if (rss->p_remain == 0) {
// move to the next block
++rss->p_block;
if (rss->p_block < rss->tgt->count) {
rss->p_remain = (rss->tgt->pos[rss->p_block * 2 + 1] -
rss->tgt->pos[rss->p_block * 2]) * BLOCKSIZE;
if (!check_lseek(rss->fd, (off64_t)rss->tgt->pos[rss->p_block*2] * BLOCKSIZE,
SEEK_SET)) {
break;
}
} else {
// we can't write any more; return how many bytes have
// been written so far.
break;
}
}
}
return written;
}
// All of the data for all the 'new' transfers is contained in one
// file in the update package, concatenated together in the order in
// which transfers.list will need it. We want to stream it out of the
// archive (it's compressed) without writing it to a temp file, but we
// can't write each section until it's that transfer's turn to go.
//
// To achieve this, we expand the new data from the archive in a
// background thread, and block that threads 'receive uncompressed
// data' function until the main thread has reached a point where we
// want some new data to be written. We signal the background thread
// with the destination for the data and block the main thread,
// waiting for the background thread to complete writing that section.
// Then it signals the main thread to wake up and goes back to
// blocking waiting for a transfer.
//
// NewThreadInfo is the struct used to pass information back and forth
// between the two threads. When the main thread wants some data
// written, it sets rss to the destination location and signals the
// condition. When the background thread is done writing, it clears
// rss and signals the condition again.
typedef struct {
ZipArchive* za;
const ZipEntry* entry;
RangeSinkState* rss;
pthread_mutex_t mu;
pthread_cond_t cv;
} NewThreadInfo;
static bool receive_new_data(const unsigned char* data, int size, void* cookie) {
NewThreadInfo* nti = (NewThreadInfo*) cookie;
while (size > 0) {
// Wait for nti->rss to be non-NULL, indicating some of this
// data is wanted.
pthread_mutex_lock(&nti->mu);
while (nti->rss == NULL) {
pthread_cond_wait(&nti->cv, &nti->mu);
}
pthread_mutex_unlock(&nti->mu);
// At this point nti->rss is set, and we own it. The main
// thread is waiting for it to disappear from nti.
ssize_t written = RangeSinkWrite(data, size, nti->rss);
data += written;
size -= written;
if (nti->rss->p_block == nti->rss->tgt->count) {
// we have written all the bytes desired by this rss.
pthread_mutex_lock(&nti->mu);
nti->rss = NULL;
pthread_cond_broadcast(&nti->cv);
pthread_mutex_unlock(&nti->mu);
}
}
return true;
}
static void* unzip_new_data(void* cookie) {
NewThreadInfo* nti = (NewThreadInfo*) cookie;
mzProcessZipEntryContents(nti->za, nti->entry, receive_new_data, nti);
return NULL;
}
static int ReadBlocks(RangeSet* src, uint8_t* buffer, int fd) {
size_t p = 0;
if (!src || !buffer) {
return -1;
}
for (size_t i = 0; i < src->count; ++i) {
if (!check_lseek(fd, (off64_t) src->pos[i * 2] * BLOCKSIZE, SEEK_SET)) {
return -1;
}
size_t size = (src->pos[i * 2 + 1] - src->pos[i * 2]) * BLOCKSIZE;
if (read_all(fd, buffer + p, size) == -1) {
return -1;
}
p += size;
}
return 0;
}
static int WriteBlocks(RangeSet* tgt, uint8_t* buffer, int fd) {
size_t p = 0;
if (!tgt || !buffer) {
return -1;
}
for (size_t i = 0; i < tgt->count; ++i) {
if (!check_lseek(fd, (off64_t) tgt->pos[i * 2] * BLOCKSIZE, SEEK_SET)) {
return -1;
}
size_t size = (tgt->pos[i * 2 + 1] - tgt->pos[i * 2]) * BLOCKSIZE;
if (write_all(fd, buffer + p, size) == -1) {
return -1;
}
p += size;
}
return 0;
}
// Do a source/target load for move/bsdiff/imgdiff in version 1.
// 'wordsave' is the save_ptr of a strtok_r()-in-progress. We expect
// to parse the remainder of the string as:
//
// <src_range> <tgt_range>
//
// The source range is loaded into the provided buffer, reallocating
// it to make it larger if necessary. The target ranges are returned
// in *tgt, if tgt is non-NULL.
static int LoadSrcTgtVersion1(char** wordsave, RangeSet** tgt, int* src_blocks,
uint8_t** buffer, size_t* buffer_alloc, int fd) {
char* word;
int rc;
word = strtok_r(NULL, " ", wordsave);
RangeSet* src = parse_range(word);
if (tgt != NULL) {
word = strtok_r(NULL, " ", wordsave);
*tgt = parse_range(word);
}
allocate(src->size * BLOCKSIZE, buffer, buffer_alloc);
rc = ReadBlocks(src, *buffer, fd);
*src_blocks = src->size;
free(src);
return rc;
}
static int VerifyBlocks(const char *expected, const uint8_t *buffer,
size_t blocks, bool printerror) {
int rc = -1;
uint8_t digest[SHA_DIGEST_SIZE];
if (!expected || !buffer) {
return rc;
}
SHA_hash(buffer, blocks * BLOCKSIZE, digest);
std::string hexdigest = print_sha1(digest);
rc = hexdigest != std::string(expected);
if (rc != 0 && printerror) {
fprintf(stderr, "failed to verify blocks (expected %s, read %s)\n",
expected, hexdigest.c_str());
}
return rc;
}
static std::string GetStashFileName(const std::string& base, const std::string id,
const std::string postfix) {
if (base.empty()) {
return "";
}
std::string fn(STASH_DIRECTORY_BASE);
fn += "/" + base + "/" + id + postfix;
return fn;
}
typedef void (*StashCallback)(const std::string&, void*);
// Does a best effort enumeration of stash files. Ignores possible non-file
// items in the stash directory and continues despite of errors. Calls the
// 'callback' function for each file and passes 'data' to the function as a
// parameter.
static void EnumerateStash(const std::string& dirname, StashCallback callback, void* data) {
if (dirname.empty() || callback == NULL) {
return;
}
std::unique_ptr<DIR, int(*)(DIR*)> directory(opendir(dirname.c_str()), closedir);
if (directory == NULL) {
if (errno != ENOENT) {
fprintf(stderr, "opendir \"%s\" failed: %s\n", dirname.c_str(), strerror(errno));
}
return;
}
struct dirent* item;
while ((item = readdir(directory.get())) != NULL) {
if (item->d_type != DT_REG) {
continue;
}
std::string fn = dirname + "/" + std::string(item->d_name);
callback(fn, data);
}
}
static void UpdateFileSize(const std::string& fn, void* data) {
if (fn.empty() || !data) {
return;
}
struct stat st;
if (stat(fn.c_str(), &st) == -1) {
fprintf(stderr, "stat \"%s\" failed: %s\n", fn.c_str(), strerror(errno));
return;
}
int* size = reinterpret_cast<int*>(data);
*size += st.st_size;
}
// Deletes the stash directory and all files in it. Assumes that it only
// contains files. There is nothing we can do about unlikely, but possible
// errors, so they are merely logged.
static void DeleteFile(const std::string& fn, void* data) {
if (!fn.empty()) {
fprintf(stderr, "deleting %s\n", fn.c_str());
if (unlink(fn.c_str()) == -1 && errno != ENOENT) {
fprintf(stderr, "unlink \"%s\" failed: %s\n", fn.c_str(), strerror(errno));
}
}
}
static void DeletePartial(const std::string& fn, void* data) {
if (android::base::EndsWith(fn, ".partial")) {
DeleteFile(fn, data);
}
}
static void DeleteStash(const std::string& base) {
if (base.empty()) {
return;
}
fprintf(stderr, "deleting stash %s\n", base.c_str());
std::string dirname = GetStashFileName(base, "", "");
EnumerateStash(dirname, DeleteFile, NULL);
if (rmdir(dirname.c_str()) == -1) {
if (errno != ENOENT && errno != ENOTDIR) {
fprintf(stderr, "rmdir \"%s\" failed: %s\n", dirname.c_str(), strerror(errno));
}
}
}
static int LoadStash(const std::string& base, const char* id, int verify, int* blocks,
uint8_t** buffer, size_t* buffer_alloc, bool printnoent) {
std::string fn;
int blockcount = 0;
int fd = -1;
int rc = -1;
int res;
struct stat st;
if (base.empty() || !id || !buffer || !buffer_alloc) {
goto lsout;
}
if (!blocks) {
blocks = &blockcount;
}
fn = GetStashFileName(base, std::string(id), "");
res = stat(fn.c_str(), &st);
if (res == -1) {
if (errno != ENOENT || printnoent) {
fprintf(stderr, "stat \"%s\" failed: %s\n", fn.c_str(), strerror(errno));
}
goto lsout;
}
fprintf(stderr, " loading %s\n", fn.c_str());
if ((st.st_size % BLOCKSIZE) != 0) {
fprintf(stderr, "%s size %" PRId64 " not multiple of block size %d",
fn.c_str(), static_cast<int64_t>(st.st_size), BLOCKSIZE);
goto lsout;
}
fd = TEMP_FAILURE_RETRY(open(fn.c_str(), O_RDONLY));
if (fd == -1) {
fprintf(stderr, "open \"%s\" failed: %s\n", fn.c_str(), strerror(errno));
goto lsout;
}
allocate(st.st_size, buffer, buffer_alloc);
if (read_all(fd, *buffer, st.st_size) == -1) {
goto lsout;
}
*blocks = st.st_size / BLOCKSIZE;
if (verify && VerifyBlocks(id, *buffer, *blocks, true) != 0) {
fprintf(stderr, "unexpected contents in %s\n", fn.c_str());
DeleteFile(fn, NULL);
goto lsout;
}
rc = 0;
lsout:
if (fd != -1) {
close(fd);
}
return rc;
}
static int WriteStash(const std::string& base, const char* id, int blocks,
uint8_t* buffer, bool checkspace, int *exists) {
std::string fn;
std::string cn;
std::string dname;
int fd = -1;
int rc = -1;
int dfd = -1;
int res;
struct stat st;
if (base.empty() || buffer == NULL) {
goto wsout;
}
if (checkspace && CacheSizeCheck(blocks * BLOCKSIZE) != 0) {
fprintf(stderr, "not enough space to write stash\n");
goto wsout;
}
fn = GetStashFileName(base, std::string(id), ".partial");
cn = GetStashFileName(base, std::string(id), "");
if (exists) {
res = stat(cn.c_str(), &st);
if (res == 0) {
// The file already exists and since the name is the hash of the contents,
// it's safe to assume the contents are identical (accidental hash collisions
// are unlikely)
fprintf(stderr, " skipping %d existing blocks in %s\n", blocks, cn.c_str());
*exists = 1;
rc = 0;
goto wsout;
}
*exists = 0;
}
fprintf(stderr, " writing %d blocks to %s\n", blocks, cn.c_str());
fd = TEMP_FAILURE_RETRY(open(fn.c_str(), O_WRONLY | O_CREAT | O_TRUNC, STASH_FILE_MODE));
if (fd == -1) {
fprintf(stderr, "failed to create \"%s\": %s\n", fn.c_str(), strerror(errno));
goto wsout;
}
if (write_all(fd, buffer, blocks * BLOCKSIZE) == -1) {
goto wsout;
}
if (fsync(fd) == -1) {
fprintf(stderr, "fsync \"%s\" failed: %s\n", fn.c_str(), strerror(errno));
goto wsout;
}
if (rename(fn.c_str(), cn.c_str()) == -1) {
fprintf(stderr, "rename(\"%s\", \"%s\") failed: %s\n", fn.c_str(), cn.c_str(),
strerror(errno));
goto wsout;
}
dname = GetStashFileName(base, "", "");
dfd = TEMP_FAILURE_RETRY(open(dname.c_str(), O_RDONLY | O_DIRECTORY));
if (dfd == -1) {
fprintf(stderr, "failed to open \"%s\" failed: %s\n", dname.c_str(), strerror(errno));
goto wsout;
}
if (fsync(dfd) == -1) {
fprintf(stderr, "fsync \"%s\" failed: %s\n", dname.c_str(), strerror(errno));
goto wsout;
}
rc = 0;
wsout:
if (fd != -1) {
close(fd);
}
if (dfd != -1) {
close(dfd);
}
return rc;
}
// Creates a directory for storing stash files and checks if the /cache partition
// hash enough space for the expected amount of blocks we need to store. Returns
// >0 if we created the directory, zero if it existed already, and <0 of failure.
static int CreateStash(State* state, int maxblocks, const char* blockdev,
std::string& base) {
if (blockdev == NULL) {
return -1;
}
// Stash directory should be different for each partition to avoid conflicts
// when updating multiple partitions at the same time, so we use the hash of
// the block device name as the base directory
SHA_CTX ctx;
SHA_init(&ctx);
SHA_update(&ctx, blockdev, strlen(blockdev));
const uint8_t* digest = SHA_final(&ctx);
base = print_sha1(digest);
std::string dirname = GetStashFileName(base, "", "");
struct stat st;
int res = stat(dirname.c_str(), &st);
if (res == -1 && errno != ENOENT) {
ErrorAbort(state, "stat \"%s\" failed: %s\n", dirname.c_str(), strerror(errno));
return -1;
} else if (res != 0) {
fprintf(stderr, "creating stash %s\n", dirname.c_str());
res = mkdir(dirname.c_str(), STASH_DIRECTORY_MODE);
if (res != 0) {
ErrorAbort(state, "mkdir \"%s\" failed: %s\n", dirname.c_str(), strerror(errno));
return -1;
}
if (CacheSizeCheck(maxblocks * BLOCKSIZE) != 0) {
ErrorAbort(state, "not enough space for stash\n");
return -1;
}
return 1; // Created directory
}
fprintf(stderr, "using existing stash %s\n", dirname.c_str());
// If the directory already exists, calculate the space already allocated to
// stash files and check if there's enough for all required blocks. Delete any
// partially completed stash files first.
EnumerateStash(dirname, DeletePartial, NULL);
int size = 0;
EnumerateStash(dirname, UpdateFileSize, &size);
size = (maxblocks * BLOCKSIZE) - size;
if (size > 0 && CacheSizeCheck(size) != 0) {
ErrorAbort(state, "not enough space for stash (%d more needed)\n", size);
return -1;
}
return 0; // Using existing directory
}
static int SaveStash(const std::string& base, char** wordsave, uint8_t** buffer,
size_t* buffer_alloc, int fd, bool usehash) {
if (!wordsave || !buffer || !buffer_alloc) {
return -1;
}
char *id = strtok_r(NULL, " ", wordsave);
if (id == NULL) {
fprintf(stderr, "missing id field in stash command\n");
return -1;
}
int blocks = 0;
if (usehash && LoadStash(base, id, 1, &blocks, buffer, buffer_alloc, false) == 0) {
// Stash file already exists and has expected contents. Do not
// read from source again, as the source may have been already
// overwritten during a previous attempt.
return 0;
}
if (LoadSrcTgtVersion1(wordsave, NULL, &blocks, buffer, buffer_alloc, fd) == -1) {
return -1;
}
if (usehash && VerifyBlocks(id, *buffer, blocks, true) != 0) {
// Source blocks have unexpected contents. If we actually need this
// data later, this is an unrecoverable error. However, the command
// that uses the data may have already completed previously, so the
// possible failure will occur during source block verification.
fprintf(stderr, "failed to load source blocks for stash %s\n", id);
return 0;
}
fprintf(stderr, "stashing %d blocks to %s\n", blocks, id);
return WriteStash(base, id, blocks, *buffer, false, NULL);
}
static int FreeStash(const std::string& base, const char* id) {
if (base.empty() || id == NULL) {
return -1;
}
std::string fn = GetStashFileName(base, std::string(id), "");
DeleteFile(fn, NULL);
return 0;
}
static void MoveRange(uint8_t* dest, RangeSet* locs, const uint8_t* source) {
// source contains packed data, which we want to move to the
// locations given in *locs in the dest buffer. source and dest
// may be the same buffer.
int start = locs->size;
int i;
for (i = locs->count-1; i >= 0; --i) {
int blocks = locs->pos[i*2+1] - locs->pos[i*2];
start -= blocks;
memmove(dest + (locs->pos[i*2] * BLOCKSIZE), source + (start * BLOCKSIZE),
blocks * BLOCKSIZE);
}
}
// Do a source/target load for move/bsdiff/imgdiff in version 2.
// 'wordsave' is the save_ptr of a strtok_r()-in-progress. We expect
// to parse the remainder of the string as one of:
//
// <tgt_range> <src_block_count> <src_range>
// (loads data from source image only)
//
// <tgt_range> <src_block_count> - <[stash_id:stash_range] ...>
// (loads data from stashes only)
//
// <tgt_range> <src_block_count> <src_range> <src_loc> <[stash_id:stash_range] ...>
// (loads data from both source image and stashes)
//
// On return, buffer is filled with the loaded source data (rearranged
// and combined with stashed data as necessary). buffer may be
// reallocated if needed to accommodate the source data. *tgt is the
// target RangeSet. Any stashes required are loaded using LoadStash.
static int LoadSrcTgtVersion2(char** wordsave, RangeSet** tgt, int* src_blocks,
uint8_t** buffer, size_t* buffer_alloc, int fd,
const std::string& stashbase, bool* overlap) {
char* word;
char* colonsave;
char* colon;
int res;
RangeSet* locs;
size_t stashalloc = 0;
uint8_t* stash = NULL;
if (tgt != NULL) {
word = strtok_r(NULL, " ", wordsave);
*tgt = parse_range(word);
}
word = strtok_r(NULL, " ", wordsave);
*src_blocks = strtol(word, NULL, 0);
allocate(*src_blocks * BLOCKSIZE, buffer, buffer_alloc);
word = strtok_r(NULL, " ", wordsave);
if (word[0] == '-' && word[1] == '\0') {
// no source ranges, only stashes
} else {
RangeSet* src = parse_range(word);
res = ReadBlocks(src, *buffer, fd);
if (overlap && tgt) {
*overlap = range_overlaps(*src, **tgt);
}
free(src);
if (res == -1) {
return -1;
}
word = strtok_r(NULL, " ", wordsave);
if (word == NULL) {
// no stashes, only source range
return 0;
}
locs = parse_range(word);
MoveRange(*buffer, locs, *buffer);
free(locs);
}
while ((word = strtok_r(NULL, " ", wordsave)) != NULL) {
// Each word is a an index into the stash table, a colon, and
// then a rangeset describing where in the source block that
// stashed data should go.
colonsave = NULL;
colon = strtok_r(word, ":", &colonsave);
res = LoadStash(stashbase, colon, 0, NULL, &stash, &stashalloc, true);
if (res == -1) {
// These source blocks will fail verification if used later, but we
// will let the caller decide if this is a fatal failure
fprintf(stderr, "failed to load stash %s\n", colon);
continue;
}
colon = strtok_r(NULL, ":", &colonsave);
locs = parse_range(colon);
MoveRange(*buffer, locs, stash);
free(locs);
}
if (stash) {
free(stash);
}
return 0;
}
// Parameters for transfer list command functions
typedef struct {
char* cmdname;
char* cpos;
char* freestash;
std::string stashbase;
bool canwrite;
int createdstash;
int fd;
int foundwrites;
bool isunresumable;
int version;
int written;
NewThreadInfo nti;
pthread_t thread;
size_t bufsize;
uint8_t* buffer;
uint8_t* patch_start;
} CommandParameters;
// Do a source/target load for move/bsdiff/imgdiff in version 3.
//
// Parameters are the same as for LoadSrcTgtVersion2, except for 'onehash', which
// tells the function whether to expect separate source and targe block hashes, or
// if they are both the same and only one hash should be expected, and
// 'isunresumable', which receives a non-zero value if block verification fails in
// a way that the update cannot be resumed anymore.
//
// If the function is unable to load the necessary blocks or their contents don't
// match the hashes, the return value is -1 and the command should be aborted.
//
// If the return value is 1, the command has already been completed according to
// the contents of the target blocks, and should not be performed again.
//
// If the return value is 0, source blocks have expected content and the command
// can be performed.
static int LoadSrcTgtVersion3(CommandParameters* params, RangeSet** tgt, int* src_blocks,
int onehash, bool* overlap) {
char* srchash = NULL;
char* tgthash = NULL;
int stash_exists = 0;
int rc = -1;
uint8_t* tgtbuffer = NULL;
if (!params|| !tgt || !src_blocks || !overlap) {
goto v3out;
}
srchash = strtok_r(NULL, " ", &params->cpos);
if (srchash == NULL) {
fprintf(stderr, "missing source hash\n");
goto v3out;
}
if (onehash) {
tgthash = srchash;
} else {
tgthash = strtok_r(NULL, " ", &params->cpos);
if (tgthash == NULL) {
fprintf(stderr, "missing target hash\n");
goto v3out;
}
}
if (LoadSrcTgtVersion2(&params->cpos, tgt, src_blocks, &params->buffer, &params->bufsize,
params->fd, params->stashbase, overlap) == -1) {
goto v3out;
}
tgtbuffer = (uint8_t*) malloc((*tgt)->size * BLOCKSIZE);
if (tgtbuffer == NULL) {
fprintf(stderr, "failed to allocate %d bytes\n", (*tgt)->size * BLOCKSIZE);
goto v3out;
}
if (ReadBlocks(*tgt, tgtbuffer, params->fd) == -1) {
goto v3out;
}
if (VerifyBlocks(tgthash, tgtbuffer, (*tgt)->size, false) == 0) {
// Target blocks already have expected content, command should be skipped
rc = 1;
goto v3out;
}
if (VerifyBlocks(srchash, params->buffer, *src_blocks, true) == 0) {
// If source and target blocks overlap, stash the source blocks so we can
// resume from possible write errors
if (*overlap) {
fprintf(stderr, "stashing %d overlapping blocks to %s\n", *src_blocks,
srchash);
if (WriteStash(params->stashbase, srchash, *src_blocks, params->buffer, true,
&stash_exists) != 0) {
fprintf(stderr, "failed to stash overlapping source blocks\n");
goto v3out;
}
// Can be deleted when the write has completed
if (!stash_exists) {
params->freestash = srchash;
}
}
// Source blocks have expected content, command can proceed
rc = 0;
goto v3out;
}
if (*overlap && LoadStash(params->stashbase, srchash, 1, NULL, &params->buffer,
&params->bufsize, true) == 0) {
// Overlapping source blocks were previously stashed, command can proceed.
// We are recovering from an interrupted command, so we don't know if the
// stash can safely be deleted after this command.
rc = 0;
goto v3out;
}
// Valid source data not available, update cannot be resumed
fprintf(stderr, "partition has unexpected contents\n");
params->isunresumable = true;
v3out:
if (tgtbuffer) {
free(tgtbuffer);
}
return rc;
}
static int PerformCommandMove(CommandParameters* params) {
int blocks = 0;
bool overlap = false;
int rc = -1;
int status = 0;
RangeSet* tgt = NULL;
if (!params) {
goto pcmout;
}
if (params->version == 1) {
status = LoadSrcTgtVersion1(&params->cpos, &tgt, &blocks, &params->buffer,
&params->bufsize, params->fd);
} else if (params->version == 2) {
status = LoadSrcTgtVersion2(&params->cpos, &tgt, &blocks, &params->buffer,
&params->bufsize, params->fd, params->stashbase, NULL);
} else if (params->version >= 3) {
status = LoadSrcTgtVersion3(params, &tgt, &blocks, 1, &overlap);
}
if (status == -1) {
fprintf(stderr, "failed to read blocks for move\n");
goto pcmout;
}
if (status == 0) {
params->foundwrites = 1;
} else if (params->foundwrites) {
fprintf(stderr, "warning: commands executed out of order [%s]\n", params->cmdname);
}
if (params->canwrite) {
if (status == 0) {
fprintf(stderr, " moving %d blocks\n", blocks);
if (WriteBlocks(tgt, params->buffer, params->fd) == -1) {
goto pcmout;
}
} else {
fprintf(stderr, "skipping %d already moved blocks\n", blocks);
}
}
if (params->freestash) {
FreeStash(params->stashbase, params->freestash);
params->freestash = NULL;
}
params->written += tgt->size;
rc = 0;
pcmout:
if (tgt) {
free(tgt);
}
return rc;
}
static int PerformCommandStash(CommandParameters* params) {
if (!params) {
return -1;
}
return SaveStash(params->stashbase, &params->cpos, &params->buffer, &params->bufsize,
params->fd, (params->version >= 3));
}
static int PerformCommandFree(CommandParameters* params) {
if (!params) {
return -1;
}
if (params->createdstash || params->canwrite) {
return FreeStash(params->stashbase, params->cpos);
}
return 0;
}
static int PerformCommandZero(CommandParameters* params) {
char* range = NULL;
int rc = -1;
RangeSet* tgt = NULL;
if (!params) {
goto pczout;
}
range = strtok_r(NULL, " ", &params->cpos);
if (range == NULL) {
fprintf(stderr, "missing target blocks for zero\n");
goto pczout;
}
tgt = parse_range(range);
fprintf(stderr, " zeroing %d blocks\n", tgt->size);
allocate(BLOCKSIZE, &params->buffer, &params->bufsize);
memset(params->buffer, 0, BLOCKSIZE);
if (params->canwrite) {
for (size_t i = 0; i < tgt->count; ++i) {
if (!check_lseek(params->fd, (off64_t) tgt->pos[i * 2] * BLOCKSIZE, SEEK_SET)) {
goto pczout;
}
for (size_t j = tgt->pos[i * 2]; j < tgt->pos[i * 2 + 1]; ++j) {
if (write_all(params->fd, params->buffer, BLOCKSIZE) == -1) {
goto pczout;
}
}
}
}
if (params->cmdname[0] == 'z') {
// Update only for the zero command, as the erase command will call
// this if DEBUG_ERASE is defined.
params->written += tgt->size;
}
rc = 0;
pczout:
if (tgt) {
free(tgt);
}
return rc;
}
static int PerformCommandNew(CommandParameters* params) {
char* range = NULL;
int rc = -1;
RangeSet* tgt = NULL;
RangeSinkState rss;
if (!params) {
goto pcnout;
}
range = strtok_r(NULL, " ", &params->cpos);
if (range == NULL) {
goto pcnout;
}
tgt = parse_range(range);
if (params->canwrite) {
fprintf(stderr, " writing %d blocks of new data\n", tgt->size);
rss.fd = params->fd;
rss.tgt = tgt;
rss.p_block = 0;
rss.p_remain = (tgt->pos[1] - tgt->pos[0]) * BLOCKSIZE;
if (!check_lseek(params->fd, (off64_t) tgt->pos[0] * BLOCKSIZE, SEEK_SET)) {
goto pcnout;
}
pthread_mutex_lock(&params->nti.mu);
params->nti.rss = &rss;
pthread_cond_broadcast(&params->nti.cv);
while (params->nti.rss) {
pthread_cond_wait(&params->nti.cv, &params->nti.mu);
}
pthread_mutex_unlock(&params->nti.mu);
}
params->written += tgt->size;
rc = 0;
pcnout:
if (tgt) {
free(tgt);
}
return rc;
}
static int PerformCommandDiff(CommandParameters* params) {
char* logparams = NULL;
char* value = NULL;
int blocks = 0;
bool overlap = false;
int rc = -1;
int status = 0;
RangeSet* tgt = NULL;
RangeSinkState rss;
size_t len = 0;
size_t offset = 0;
Value patch_value;
if (!params) {
goto pcdout;
}
logparams = strdup(params->cpos);
value = strtok_r(NULL, " ", &params->cpos);
if (value == NULL) {
fprintf(stderr, "missing patch offset for %s\n", params->cmdname);
goto pcdout;
}
offset = strtoul(value, NULL, 0);
value = strtok_r(NULL, " ", &params->cpos);
if (value == NULL) {
fprintf(stderr, "missing patch length for %s\n", params->cmdname);
goto pcdout;
}
len = strtoul(value, NULL, 0);
if (params->version == 1) {
status = LoadSrcTgtVersion1(&params->cpos, &tgt, &blocks, &params->buffer,
&params->bufsize, params->fd);
} else if (params->version == 2) {
status = LoadSrcTgtVersion2(&params->cpos, &tgt, &blocks, &params->buffer,
&params->bufsize, params->fd, params->stashbase, NULL);
} else if (params->version >= 3) {
status = LoadSrcTgtVersion3(params, &tgt, &blocks, 0, &overlap);
}
if (status == -1) {
fprintf(stderr, "failed to read blocks for diff\n");
goto pcdout;
}
if (status == 0) {
params->foundwrites = 1;
} else if (params->foundwrites) {
fprintf(stderr, "warning: commands executed out of order [%s]\n", params->cmdname);
}
if (params->canwrite) {
if (status == 0) {
fprintf(stderr, "patching %d blocks to %d\n", blocks, tgt->size);
patch_value.type = VAL_BLOB;
patch_value.size = len;
patch_value.data = (char*) (params->patch_start + offset);
rss.fd = params->fd;
rss.tgt = tgt;
rss.p_block = 0;
rss.p_remain = (tgt->pos[1] - tgt->pos[0]) * BLOCKSIZE;
if (!check_lseek(params->fd, (off64_t) tgt->pos[0] * BLOCKSIZE, SEEK_SET)) {
goto pcdout;
}
if (params->cmdname[0] == 'i') { // imgdiff
ApplyImagePatch(params->buffer, blocks * BLOCKSIZE, &patch_value,
&RangeSinkWrite, &rss, NULL, NULL);
} else {
ApplyBSDiffPatch(params->buffer, blocks * BLOCKSIZE, &patch_value,
0, &RangeSinkWrite, &rss, NULL);
}
// We expect the output of the patcher to fill the tgt ranges exactly.
if (rss.p_block != tgt->count || rss.p_remain != 0) {
fprintf(stderr, "range sink underrun?\n");
}
} else {
fprintf(stderr, "skipping %d blocks already patched to %d [%s]\n",
blocks, tgt->size, logparams);
}
}
if (params->freestash) {
FreeStash(params->stashbase, params->freestash);
params->freestash = NULL;
}
params->written += tgt->size;
rc = 0;
pcdout:
if (logparams) {
free(logparams);
}
if (tgt) {
free(tgt);
}
return rc;
}
static int PerformCommandErase(CommandParameters* params) {
char* range = NULL;
int rc = -1;
RangeSet* tgt = NULL;
struct stat st;
uint64_t blocks[2];
if (DEBUG_ERASE) {
return PerformCommandZero(params);
}
if (!params) {
goto pceout;
}
if (fstat(params->fd, &st) == -1) {
fprintf(stderr, "failed to fstat device to erase: %s\n", strerror(errno));
goto pceout;
}
if (!S_ISBLK(st.st_mode)) {
fprintf(stderr, "not a block device; skipping erase\n");
goto pceout;
}
range = strtok_r(NULL, " ", &params->cpos);
if (range == NULL) {
fprintf(stderr, "missing target blocks for erase\n");
goto pceout;
}
tgt = parse_range(range);
if (params->canwrite) {
fprintf(stderr, " erasing %d blocks\n", tgt->size);
for (size_t i = 0; i < tgt->count; ++i) {
// offset in bytes
blocks[0] = tgt->pos[i * 2] * (uint64_t) BLOCKSIZE;
// length in bytes
blocks[1] = (tgt->pos[i * 2 + 1] - tgt->pos[i * 2]) * (uint64_t) BLOCKSIZE;
if (ioctl(params->fd, BLKDISCARD, &blocks) == -1) {
fprintf(stderr, "BLKDISCARD ioctl failed: %s\n", strerror(errno));
goto pceout;
}
}
}
rc = 0;
pceout:
if (tgt) {
free(tgt);
}
return rc;
}
// Definitions for transfer list command functions
typedef int (*CommandFunction)(CommandParameters*);
typedef struct {
const char* name;
CommandFunction f;
} Command;
// CompareCommands and CompareCommandNames are for the hash table
static int CompareCommands(const void* c1, const void* c2) {
return strcmp(((const Command*) c1)->name, ((const Command*) c2)->name);
}
static int CompareCommandNames(const void* c1, const void* c2) {
return strcmp(((const Command*) c1)->name, (const char*) c2);
}
// HashString is used to hash command names for the hash table
static unsigned int HashString(const char *s) {
unsigned int hash = 0;
if (s) {
while (*s) {
hash = hash * 33 + *s++;
}
}
return hash;
}
// args:
// - block device (or file) to modify in-place
// - transfer list (blob)
// - new data stream (filename within package.zip)
// - patch stream (filename within package.zip, must be uncompressed)
static Value* PerformBlockImageUpdate(const char* name, State* state, int argc, Expr* argv[],
const Command* commands, int cmdcount, bool dryrun) {
char* line = NULL;
char* linesave = NULL;
char* logcmd = NULL;
char* transfer_list = NULL;
CommandParameters params;
const Command* cmd = NULL;
const ZipEntry* new_entry = NULL;
const ZipEntry* patch_entry = NULL;
FILE* cmd_pipe = NULL;
HashTable* cmdht = NULL;
int i;
int res;
int rc = -1;
int stash_max_blocks = 0;
int total_blocks = 0;
pthread_attr_t attr;
unsigned int cmdhash;
UpdaterInfo* ui = NULL;
Value* blockdev_filename = NULL;
Value* new_data_fn = NULL;
Value* patch_data_fn = NULL;
Value* transfer_list_value = NULL;
ZipArchive* za = NULL;
memset(&params, 0, sizeof(params));
params.canwrite = !dryrun;
fprintf(stderr, "performing %s\n", dryrun ? "verification" : "update");
if (ReadValueArgs(state, argv, 4, &blockdev_filename, &transfer_list_value,
&new_data_fn, &patch_data_fn) < 0) {
goto pbiudone;
}
if (blockdev_filename->type != VAL_STRING) {
ErrorAbort(state, "blockdev_filename argument to %s must be string", name);
goto pbiudone;
}
if (transfer_list_value->type != VAL_BLOB) {
ErrorAbort(state, "transfer_list argument to %s must be blob", name);
goto pbiudone;
}
if (new_data_fn->type != VAL_STRING) {
ErrorAbort(state, "new_data_fn argument to %s must be string", name);
goto pbiudone;
}
if (patch_data_fn->type != VAL_STRING) {
ErrorAbort(state, "patch_data_fn argument to %s must be string", name);
goto pbiudone;
}
ui = (UpdaterInfo*) state->cookie;
if (ui == NULL) {
goto pbiudone;
}
cmd_pipe = ui->cmd_pipe;
za = ui->package_zip;
if (cmd_pipe == NULL || za == NULL) {
goto pbiudone;
}
patch_entry = mzFindZipEntry(za, patch_data_fn->data);
if (patch_entry == NULL) {
fprintf(stderr, "%s(): no file \"%s\" in package", name, patch_data_fn->data);
goto pbiudone;
}
params.patch_start = ui->package_zip_addr + mzGetZipEntryOffset(patch_entry);
new_entry = mzFindZipEntry(za, new_data_fn->data);
if (new_entry == NULL) {
fprintf(stderr, "%s(): no file \"%s\" in package", name, new_data_fn->data);
goto pbiudone;
}
params.fd = TEMP_FAILURE_RETRY(open(blockdev_filename->data, O_RDWR));
if (params.fd == -1) {
fprintf(stderr, "open \"%s\" failed: %s\n", blockdev_filename->data, strerror(errno));
goto pbiudone;
}
if (params.canwrite) {
params.nti.za = za;
params.nti.entry = new_entry;
pthread_mutex_init(&params.nti.mu, NULL);
pthread_cond_init(&params.nti.cv, NULL);
pthread_attr_init(&attr);
pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_JOINABLE);
int error = pthread_create(&params.thread, &attr, unzip_new_data, &params.nti);
if (error != 0) {
fprintf(stderr, "pthread_create failed: %s\n", strerror(error));
goto pbiudone;
}
}
// The data in transfer_list_value is not necessarily null-terminated, so we need
// to copy it to a new buffer and add the null that strtok_r will need.
transfer_list = reinterpret_cast<char*>(malloc(transfer_list_value->size + 1));
if (transfer_list == NULL) {
fprintf(stderr, "failed to allocate %zd bytes for transfer list\n",
transfer_list_value->size + 1);
goto pbiudone;
}
memcpy(transfer_list, transfer_list_value->data, transfer_list_value->size);
transfer_list[transfer_list_value->size] = '\0';
// First line in transfer list is the version number
line = strtok_r(transfer_list, "\n", &linesave);
params.version = strtol(line, NULL, 0);
if (params.version < 1 || params.version > 3) {
fprintf(stderr, "unexpected transfer list version [%s]\n", line);
goto pbiudone;
}
fprintf(stderr, "blockimg version is %d\n", params.version);
// Second line in transfer list is the total number of blocks we expect to write
line = strtok_r(NULL, "\n", &linesave);
total_blocks = strtol(line, NULL, 0);
if (total_blocks < 0) {
ErrorAbort(state, "unexpected block count [%s]\n", line);
goto pbiudone;
} else if (total_blocks == 0) {
rc = 0;
goto pbiudone;
}
if (params.version >= 2) {
// Third line is how many stash entries are needed simultaneously
line = strtok_r(NULL, "\n", &linesave);
fprintf(stderr, "maximum stash entries %s\n", line);
// Fourth line is the maximum number of blocks that will be stashed simultaneously
line = strtok_r(NULL, "\n", &linesave);
stash_max_blocks = strtol(line, NULL, 0);
if (stash_max_blocks < 0) {
ErrorAbort(state, "unexpected maximum stash blocks [%s]\n", line);
goto pbiudone;
}
if (stash_max_blocks >= 0) {
res = CreateStash(state, stash_max_blocks, blockdev_filename->data, params.stashbase);
if (res == -1) {
goto pbiudone;
}
params.createdstash = res;
}
}
// Build a hash table of the available commands
cmdht = mzHashTableCreate(cmdcount, NULL);
for (i = 0; i < cmdcount; ++i) {
cmdhash = HashString(commands[i].name);
mzHashTableLookup(cmdht, cmdhash, (void*) &commands[i], CompareCommands, true);
}
// Subsequent lines are all individual transfer commands
for (line = strtok_r(NULL, "\n", &linesave); line;
line = strtok_r(NULL, "\n", &linesave)) {
logcmd = strdup(line);
params.cmdname = strtok_r(line, " ", &params.cpos);
if (params.cmdname == NULL) {
fprintf(stderr, "missing command [%s]\n", line);
goto pbiudone;
}
cmdhash = HashString(params.cmdname);
cmd = (const Command*) mzHashTableLookup(cmdht, cmdhash, params.cmdname,
CompareCommandNames, false);
if (cmd == NULL) {
fprintf(stderr, "unexpected command [%s]\n", params.cmdname);
goto pbiudone;
}
if (cmd->f != NULL && cmd->f(&params) == -1) {
fprintf(stderr, "failed to execute command [%s]\n",
logcmd ? logcmd : params.cmdname);
goto pbiudone;
}
if (logcmd) {
free(logcmd);
logcmd = NULL;
}
if (params.canwrite) {
if (fsync(params.fd) == -1) {
fprintf(stderr, "fsync failed: %s\n", strerror(errno));
goto pbiudone;
}
fprintf(cmd_pipe, "set_progress %.4f\n", (double) params.written / total_blocks);
fflush(cmd_pipe);
}
}
if (params.canwrite) {
pthread_join(params.thread, NULL);
fprintf(stderr, "wrote %d blocks; expected %d\n", params.written, total_blocks);
fprintf(stderr, "max alloc needed was %zu\n", params.bufsize);
// Delete stash only after successfully completing the update, as it
// may contain blocks needed to complete the update later.
DeleteStash(params.stashbase);
} else {
fprintf(stderr, "verified partition contents; update may be resumed\n");
}
rc = 0;
pbiudone:
if (params.fd != -1) {
if (fsync(params.fd) == -1) {
fprintf(stderr, "fsync failed: %s\n", strerror(errno));
}
close(params.fd);
}
if (logcmd) {
free(logcmd);
}
if (cmdht) {
mzHashTableFree(cmdht);
}
if (params.buffer) {
free(params.buffer);
}
if (transfer_list) {
free(transfer_list);
}
if (blockdev_filename) {
FreeValue(blockdev_filename);
}
if (transfer_list_value) {
FreeValue(transfer_list_value);
}
if (new_data_fn) {
FreeValue(new_data_fn);
}
if (patch_data_fn) {
FreeValue(patch_data_fn);
}
// Only delete the stash if the update cannot be resumed, or it's
// a verification run and we created the stash.
if (params.isunresumable || (!params.canwrite && params.createdstash)) {
DeleteStash(params.stashbase);
}
return StringValue(rc == 0 ? strdup("t") : strdup(""));
}
// The transfer list is a text file containing commands to
// transfer data from one place to another on the target
// partition. We parse it and execute the commands in order:
//
// zero [rangeset]
// - fill the indicated blocks with zeros
//
// new [rangeset]
// - fill the blocks with data read from the new_data file
//
// erase [rangeset]
// - mark the given blocks as empty
//
// move <...>
// bsdiff <patchstart> <patchlen> <...>
// imgdiff <patchstart> <patchlen> <...>
// - read the source blocks, apply a patch (or not in the
// case of move), write result to target blocks. bsdiff or
// imgdiff specifies the type of patch; move means no patch
// at all.
//
// The format of <...> differs between versions 1 and 2;
// see the LoadSrcTgtVersion{1,2}() functions for a
// description of what's expected.
//
// stash <stash_id> <src_range>
// - (version 2+ only) load the given source range and stash
// the data in the given slot of the stash table.
//
// The creator of the transfer list will guarantee that no block
// is read (ie, used as the source for a patch or move) after it
// has been written.
//
// In version 2, the creator will guarantee that a given stash is
// loaded (with a stash command) before it's used in a
// move/bsdiff/imgdiff command.
//
// Within one command the source and target ranges may overlap so
// in general we need to read the entire source into memory before
// writing anything to the target blocks.
//
// All the patch data is concatenated into one patch_data file in
// the update package. It must be stored uncompressed because we
// memory-map it in directly from the archive. (Since patches are
// already compressed, we lose very little by not compressing
// their concatenation.)
//
// In version 3, commands that read data from the partition (i.e.
// move/bsdiff/imgdiff/stash) have one or more additional hashes
// before the range parameters, which are used to check if the
// command has already been completed and verify the integrity of
// the source data.
Value* BlockImageVerifyFn(const char* name, State* state, int argc, Expr* argv[]) {
// Commands which are not tested are set to NULL to skip them completely
const Command commands[] = {
{ "bsdiff", PerformCommandDiff },
{ "erase", NULL },
{ "free", PerformCommandFree },
{ "imgdiff", PerformCommandDiff },
{ "move", PerformCommandMove },
{ "new", NULL },
{ "stash", PerformCommandStash },
{ "zero", NULL }
};
// Perform a dry run without writing to test if an update can proceed
return PerformBlockImageUpdate(name, state, argc, argv, commands,
sizeof(commands) / sizeof(commands[0]), true);
}
Value* BlockImageUpdateFn(const char* name, State* state, int argc, Expr* argv[]) {
const Command commands[] = {
{ "bsdiff", PerformCommandDiff },
{ "erase", PerformCommandErase },
{ "free", PerformCommandFree },
{ "imgdiff", PerformCommandDiff },
{ "move", PerformCommandMove },
{ "new", PerformCommandNew },
{ "stash", PerformCommandStash },
{ "zero", PerformCommandZero }
};
return PerformBlockImageUpdate(name, state, argc, argv, commands,
sizeof(commands) / sizeof(commands[0]), false);
}
Value* RangeSha1Fn(const char* name, State* state, int argc, Expr* argv[]) {
Value* blockdev_filename;
Value* ranges;
const uint8_t* digest = NULL;
if (ReadValueArgs(state, argv, 2, &blockdev_filename, &ranges) < 0) {
return NULL;
}
if (blockdev_filename->type != VAL_STRING) {
ErrorAbort(state, "blockdev_filename argument to %s must be string", name);
goto done;
}
if (ranges->type != VAL_STRING) {
ErrorAbort(state, "ranges argument to %s must be string", name);
goto done;
}
int fd;
fd = open(blockdev_filename->data, O_RDWR);
if (fd < 0) {
ErrorAbort(state, "open \"%s\" failed: %s", blockdev_filename->data, strerror(errno));
goto done;
}
RangeSet* rs;
rs = parse_range(ranges->data);
uint8_t buffer[BLOCKSIZE];
SHA_CTX ctx;
SHA_init(&ctx);
for (size_t i = 0; i < rs->count; ++i) {
if (!check_lseek(fd, (off64_t)rs->pos[i*2] * BLOCKSIZE, SEEK_SET)) {
ErrorAbort(state, "failed to seek %s: %s", blockdev_filename->data,
strerror(errno));
goto done;
}
for (size_t j = rs->pos[i*2]; j < rs->pos[i*2+1]; ++j) {
if (read_all(fd, buffer, BLOCKSIZE) == -1) {
ErrorAbort(state, "failed to read %s: %s", blockdev_filename->data,
strerror(errno));
goto done;
}
SHA_update(&ctx, buffer, BLOCKSIZE);
}
}
digest = SHA_final(&ctx);
close(fd);
done:
FreeValue(blockdev_filename);
FreeValue(ranges);
if (digest == NULL) {
return StringValue(strdup(""));
} else {
return StringValue(strdup(print_sha1(digest).c_str()));
}
}
void RegisterBlockImageFunctions() {
RegisterFunction("block_image_verify", BlockImageVerifyFn);
RegisterFunction("block_image_update", BlockImageUpdateFn);
RegisterFunction("range_sha1", RangeSha1Fn);
}