From: "Darrick J. Wong" Subject: Re: [PATCH 15/34] libext2fs: support BLKZEROOUT/FALLOC_FL_ZERO_RANGE in ext2fs_zero_blocks Date: Mon, 13 Oct 2014 19:58:53 -0700 Message-ID: <20141014025853.GI12009@birch.djwong.org> References: <20140913221112.13646.3873.stgit@birch.djwong.org> <20140913221253.13646.7723.stgit@birch.djwong.org> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: linux-ext4@vger.kernel.org To: tytso@mit.edu Return-path: Received: from userp1040.oracle.com ([156.151.31.81]:45535 "EHLO userp1040.oracle.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754410AbaJNC67 (ORCPT ); Mon, 13 Oct 2014 22:58:59 -0400 Content-Disposition: inline In-Reply-To: <20140913221253.13646.7723.stgit@birch.djwong.org> Sender: linux-ext4-owner@vger.kernel.org List-ID: On Sat, Sep 13, 2014 at 03:12:53PM -0700, Darrick J. Wong wrote: > Plumb a new call into the IO manager to support translating > ext2fs_zero_blocks calls into the equivalent kernel-level BLKZEROOUT > ioctl or FALLOC_FL_ZERO_RANGE fallocate flag primitives when possible. > > Signed-off-by: Darrick J. Wong > --- > contrib/fallocate.c | 14 +++++++++ > lib/ext2fs/ext2_io.h | 7 ++++- > lib/ext2fs/io_manager.c | 11 +++++++ > lib/ext2fs/mkjournal.c | 6 ++++ > lib/ext2fs/test_io.c | 21 ++++++++++++++ > lib/ext2fs/unix_io.c | 71 +++++++++++++++++++++++++++++++++++++++++++++++ > 6 files changed, 128 insertions(+), 2 deletions(-) > > > diff --git a/contrib/fallocate.c b/contrib/fallocate.c > index 1f9b59a..01d4af7 100644 > --- a/contrib/fallocate.c > +++ b/contrib/fallocate.c > @@ -36,6 +36,8 @@ > // #include > #define FALLOC_FL_KEEP_SIZE 0x01 > #define FALLOC_FL_PUNCH_HOLE 0x02 /* de-allocates range */ > +#define FALLOC_FL_COLLAPSE_RANGE 0x08 > +#define FALLOC_FL_ZERO_RANGE 0x10 > > void usage(void) > { > @@ -95,7 +97,7 @@ int main(int argc, char **argv) > int error; > int tflag = 0; > > - while ((opt = getopt(argc, argv, "npl:o:t")) != -1) { > + while ((opt = getopt(argc, argv, "npl:o:tzc")) != -1) { > switch(opt) { > case 'n': > /* do not change filesize */ > @@ -106,6 +108,16 @@ int main(int argc, char **argv) > falloc_mode = (FALLOC_FL_PUNCH_HOLE | > FALLOC_FL_KEEP_SIZE); > break; > + case 'c': > + /* collapse range mode */ > + falloc_mode = (FALLOC_FL_COLLAPSE_RANGE | > + FALLOC_FL_KEEP_SIZE); > + break; > + case 'z': > + /* zero range mode */ > + falloc_mode = (FALLOC_FL_ZERO_RANGE | > + FALLOC_FL_KEEP_SIZE); > + break; > case 'l': > length = cvtnum(optarg); > break; > diff --git a/lib/ext2fs/ext2_io.h b/lib/ext2fs/ext2_io.h > index 4c5a5c5..1faa720 100644 > --- a/lib/ext2fs/ext2_io.h > +++ b/lib/ext2fs/ext2_io.h > @@ -93,7 +93,9 @@ struct struct_io_manager { > errcode_t (*cache_readahead)(io_channel channel, > unsigned long long block, > unsigned long long count); > - long reserved[15]; > + errcode_t (*zeroout)(io_channel channel, unsigned long long block, > + unsigned long long count); > + long reserved[14]; > }; > > #define IO_FLAG_RW 0x0001 > @@ -125,6 +127,9 @@ extern errcode_t io_channel_write_blk64(io_channel channel, > extern errcode_t io_channel_discard(io_channel channel, > unsigned long long block, > unsigned long long count); > +extern errcode_t io_channel_zeroout(io_channel channel, > + unsigned long long block, > + unsigned long long count); > extern errcode_t io_channel_alloc_buf(io_channel channel, > int count, void *ptr); > extern errcode_t io_channel_cache_readahead(io_channel io, > diff --git a/lib/ext2fs/io_manager.c b/lib/ext2fs/io_manager.c > index dc5888d..c395d61 100644 > --- a/lib/ext2fs/io_manager.c > +++ b/lib/ext2fs/io_manager.c > @@ -112,6 +112,17 @@ errcode_t io_channel_discard(io_channel channel, unsigned long long block, > return EXT2_ET_UNIMPLEMENTED; > } > > +errcode_t io_channel_zeroout(io_channel channel, unsigned long long block, > + unsigned long long count) > +{ > + EXT2_CHECK_MAGIC(channel, EXT2_ET_MAGIC_IO_CHANNEL); > + > + if (channel->manager->zeroout) > + return (channel->manager->zeroout)(channel, block, count); > + > + return EXT2_ET_UNIMPLEMENTED; > +} > + > errcode_t io_channel_alloc_buf(io_channel io, int count, void *ptr) > { > size_t size; > diff --git a/lib/ext2fs/mkjournal.c b/lib/ext2fs/mkjournal.c > index 6f3a862..5be425c 100644 > --- a/lib/ext2fs/mkjournal.c > +++ b/lib/ext2fs/mkjournal.c > @@ -164,6 +164,12 @@ errcode_t ext2fs_zero_blocks2(ext2_filsys fs, blk64_t blk, int num, > } > return 0; > } > + > + /* Try a zero out command, if supported */ > + retval = io_channel_zeroout(fs->io, blk, num); > + if (retval == 0) > + return 0; > + > /* Allocate the zeroizing buffer if necessary */ > if (!buf) { > buf = malloc(fs->blocksize * STRIDE_LENGTH); > diff --git a/lib/ext2fs/test_io.c b/lib/ext2fs/test_io.c > index b03a939..f7c50d1 100644 > --- a/lib/ext2fs/test_io.c > +++ b/lib/ext2fs/test_io.c > @@ -86,6 +86,7 @@ void (*test_io_cb_write_byte) > #define TEST_FLAG_SET_OPTION 0x20 > #define TEST_FLAG_DISCARD 0x40 > #define TEST_FLAG_READAHEAD 0x80 > +#define TEST_FLAG_ZEROOUT 0x100 > > static void test_dump_block(io_channel channel, > struct test_private_data *data, > @@ -507,6 +508,25 @@ static errcode_t test_cache_readahead(io_channel channel, > return retval; > } > > +static errcode_t test_zeroout(io_channel channel, unsigned long long block, > + unsigned long long count) > +{ > + struct test_private_data *data; > + errcode_t retval = 0; > + > + EXT2_CHECK_MAGIC(channel, EXT2_ET_MAGIC_IO_CHANNEL); > + data = (struct test_private_data *) channel->private_data; > + EXT2_CHECK_MAGIC(data, EXT2_ET_MAGIC_TEST_IO_CHANNEL); > + > + if (data->real) > + retval = io_channel_zeroout(data->real, block, count); > + if (data->flags & TEST_FLAG_ZEROOUT) > + fprintf(data->outfile, > + "Test_io: zeroout(%llu, %llu) returned %s\n", > + block, count, retval ? error_message(retval) : "OK"); > + return retval; > +} > + > static struct struct_io_manager struct_test_manager = { > .magic = EXT2_ET_MAGIC_IO_MANAGER, > .name = "Test I/O Manager", > @@ -523,6 +543,7 @@ static struct struct_io_manager struct_test_manager = { > .write_blk64 = test_write_blk64, > .discard = test_discard, > .cache_readahead = test_cache_readahead, > + .zeroout = test_zeroout, > }; > > io_manager test_io_manager = &struct_test_manager; > diff --git a/lib/ext2fs/unix_io.c b/lib/ext2fs/unix_io.c > index 189adce..20e5b64 100644 > --- a/lib/ext2fs/unix_io.c > +++ b/lib/ext2fs/unix_io.c > @@ -986,6 +986,76 @@ unimplemented: > return EXT2_ET_UNIMPLEMENTED; > } > > +#if defined(__linux__) && !defined(BLKZEROOUT) > +#define BLKZEROOUT _IO(0x12, 127) > +#endif > + > +#if defined(__linux__) && !defined(FALLOC_FL_ZERO_RANGE) > +#define FALLOC_FL_ZERO_RANGE 0x10 > +#endif > + > +static errcode_t unix_zeroout(io_channel channel, unsigned long long block, > + unsigned long long count) > +{ > + struct unix_private_data *data; > + int ret; > + > + EXT2_CHECK_MAGIC(channel, EXT2_ET_MAGIC_IO_CHANNEL); > + data = (struct unix_private_data *) channel->private_data; > + EXT2_CHECK_MAGIC(data, EXT2_ET_MAGIC_UNIX_IO_CHANNEL); > + > + if (getenv("UNIX_IO_NOZEROOUT")) > + goto unimplemented; > + > + if (channel->flags & CHANNEL_FLAGS_BLOCK_DEVICE) { > +#ifdef BLKZEROOUT > + __u64 range[2]; > + > + range[0] = (__u64)(block) * channel->block_size; > + range[1] = (__u64)(count) * channel->block_size; > + > + ret = ioctl(data->dev, BLKZEROOUT, &range); NAK this line for now. I think I've uncovered a race condition where pwrite+blkzeroout+pread returns the pwritten contents instead of zeroes as I was expecting. --D > +#else > + goto unimplemented; > +#endif > + } else { > +#if defined(HAVE_FALLOCATE) && defined(FALLOC_FL_ZERO_RANGE) > + int flag = FALLOC_FL_ZERO_RANGE; > + struct stat statbuf; > + > + /* > + * If we're trying to zero a range past the end of the file, > + * just use regular fallocate to get there, because zeroing > + * a range past EOF does not extend the file. > + */ > + ret = fstat(data->dev, &statbuf); > + if (ret) > + goto err; > + if (statbuf.st_size < (block + count) * channel->block_size) > + flag = 0; > + /* > + * If we are not on block device, try to use the zero out > + * primitive. > + */ > + ret = fallocate(data->dev, > + flag, > + (off_t)(block) * channel->block_size, > + (off_t)(count) * channel->block_size); > +#else > + goto unimplemented; > +#endif > + } > +err: > + if (ret < 0) { > + if (errno == EOPNOTSUPP) > + goto unimplemented; > + return errno; > + } > + return 0; > +unimplemented: > + return EXT2_ET_UNIMPLEMENTED; > +} > + > static struct struct_io_manager struct_unix_manager = { > .magic = EXT2_ET_MAGIC_IO_MANAGER, > .name = "Unix I/O Manager", > @@ -1002,6 +1072,7 @@ static struct struct_io_manager struct_unix_manager = { > .write_blk64 = unix_write_blk64, > .discard = unix_discard, > .cache_readahead = unix_cache_readahead, > + .zeroout = unix_zeroout, > }; > > io_manager unix_io_manager = &struct_unix_manager; > > -- > To unsubscribe from this list: send the line "unsubscribe linux-ext4" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html