|
| From: | Vladimir Sementsov-Ogievskiy |
| Subject: | Re: [PATCH v14 3/4] qcow2: add zstd cluster compression |
| Date: | Wed, 1 Apr 2020 14:21:25 +0300 |
| User-agent: | Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101 Thunderbird/60.2.1 |
01.04.2020 10:02, Denis Plotnikov wrote:
On 01.04.2020 08:49, Vladimir Sementsov-Ogievskiy wrote:31.03.2020 20:44, Denis Plotnikov wrote:zstd significantly reduces cluster compression time. It provides better compression performance maintaining the same level of the compression ratio in comparison with zlib, which, at the moment, is the only compression method available. The performance test results: Test compresses and decompresses qemu qcow2 image with just installed rhel-7.6 guest. Image cluster size: 64K. Image on disk size: 2.2G The test was conducted with brd disk to reduce the influence of disk subsystem to the test results. The results is given in seconds. compress cmd: time ./qemu-img convert -O qcow2 -c -o compression_type=[zlib|zstd] src.img [zlib|zstd]_compressed.img decompress cmd time ./qemu-img convert -O qcow2 [zlib|zstd]_compressed.img uncompressed.img compression decompression zlib zstd zlib zstd ------------------------------------------------------------ real 65.5 16.3 (-75 %) 1.9 1.6 (-16 %) user 65.0 15.8 5.3 2.5 sys 3.3 0.2 2.0 2.0 Both ZLIB and ZSTD gave the same compression ratio: 1.57 compressed image size in both cases: 1.4G Signed-off-by: Denis Plotnikov <address@hidden> QAPI part: Acked-by: Markus Armbruster <address@hidden> ---[..]+static ssize_t qcow2_zstd_compress(void *dest, size_t dest_size, + const void *src, size_t src_size) +{ + ssize_t ret; + size_t zstd_ret = 0; + ZSTD_outBuffer output = { dest, dest_size, 0 }; + ZSTD_inBuffer input = { src, src_size, 0 }; + ZSTD_CCtx *cctx = ZSTD_createCCtx(); + + if (!cctx) { + return -EIO; + } + /* + * We want to use zstd streamed interface on decompression, + * as we won't know the exact size of the compressed data. + * + * In the loop, we try to compress all the data into one zstd frame. + * ZSTD_compressStream2 potentially can finish a frame earlier + * than the full input data is consumed. That's why we are looping + * until all the input data is consumed. + */ + while (input.pos < input.size) {zstd_ret may be defined here.yep!+ /* + * ZSTD spec: "You must continue calling ZSTD_compressStream2() + * with ZSTD_e_end until it returns 0, at which point you are + * free to start a new frame". We assume that "start a new frame" + * means call ZSTD_compressStream2 in the very beginning or when + * ZSTD_compressStream2 has returned with 0. + */ + do {Hmm. Why did you decide to use nested loop? Ok, it works too.The previous condition was error prone. It led to loop ending once we finished the "first frame" but there could be a number of them.+ zstd_ret = ZSTD_compressStream2(cctx, &output, &input, ZSTD_e_end); + + if (ZSTD_isError(zstd_ret)) { + ret = -EIO; + goto out; + } + /* Dest buffer isn't big enough to store compressed content */ + if (zstd_ret > output.size - output.pos) { + ret = -ENOMEM; + goto out; + } + } while (zstd_ret); + } + /* make sure we can safely return compressed buffer size with ssize_t */ + assert(output.pos <= SSIZE_MAX); + ret = output.pos; +out: + ZSTD_freeCCtx(cctx); + return ret; +} + +/* + * qcow2_zstd_decompress() + * + * Decompress some data (not more than @src_size bytes) to produce exactly + * @dest_size bytes using zstd compression method + * + * @dest - destination buffer, @dest_size bytes + * @src - source buffer, @src_size bytes + * + * Returns: 0 on success + * -EIO on any error + */ +static ssize_t qcow2_zstd_decompress(void *dest, size_t dest_size, + const void *src, size_t src_size) +{ + size_t ret = 0;You forget to fix ret here. ret of the function should be ssize_t and for ZSTD - size_t.I decided not to use two vars since we can return only 0 or -EIO. I added an assert at the very end of the function to check it.
But size_t is unsigned type. Bad idea to assign -EIO to it. And that's why we've moved to zstd_ret in _compress
+ ZSTD_outBuffer output = { dest, dest_size, 0 }; + ZSTD_inBuffer input = { src, src_size, 0 }; + ZSTD_DCtx *dctx = ZSTD_createDCtx(); + + if (!dctx) { + return -EIO; + } + + /* + * The compressed stream from input buffer may consist from more + * than one zstd frames. So we iterate until we get a fully + * uncompressed cluster. + * from zstd docs related to ZSTD_decompressStream: + * "return : 0 when a frame is completely decoded and fully flushed" + * We suppose that this means: each time ZSTD_decompressStream reads + * only ONE full frame and return 0 if and only if that frame + * is completely decoded and flushed. Only after returning 0, + * ZSTD_decompressStream reads another ONE full frame. + */ + while (output.pos < output.size) { + size_t last_in_pos = input.pos; + size_t last_out_pos = output.pos; + ret = ZSTD_decompressStream(dctx, &output, &input); + /* + * zstd manual doesn't explicitly states what happens, + * if ZSTD_decompressStream reads the frame partially. + * But, based on our tests, if we don't fully populate + * the output and have read all the frames from the input, + * we end up with error here. + */ + if (ZSTD_isError(ret)) { + ret = -EIO; + break; + } + + /* + * As ZSTD manual is vague about what to do if it + * reads the buffer partially, we afraid of case + * when we stuck in the infinite loop, because + * output isn't full and input has read partially, so + * ZSTD_decompressStream returns > 0 waiting for + * another input chunk. So, we add a paranoid check + * that on each step the loop makes some progress. + */ + if (last_in_pos >= input.pos && + last_out_pos >= output.pos) { + ret = -EIO; + break; + } + } + /* + * Make sure that we have the frame fully flushed here + * if not, we somehow managed to get uncompressed cluster + * greater then the cluster size, possibly because of its + * damage. + */ + if (ret > 0) { + ret = -EIO; + } + + ZSTD_freeDCtx(dctx); + assert(ret == 0 || ret == -EIO); + return ret; +}[..]
-- Best regards, Vladimir
| [Prev in Thread] | Current Thread | [Next in Thread] |