1 This patch adds the --sparse-block option. Andrea Righi writes:
3 In some filesystems, typically optimized for large I/O throughputs (like
4 IBM GPFS, IBM SAN FS, or distributed filesystems in general) a lot of
5 lseek() operations can strongly impact on performances. In this cases it
6 can be helpful to enlarge the block size used to handle sparse files
7 directly from a command line parameter.
9 For example, using a sparse write size of 32KB, I've been able to
10 increase the transfer rate of an order of magnitude copying the output
11 files of scientific applications from GPFS to GPFS or GPFS to SAN FS.
15 To use this patch, run these commands for a successful build:
17 patch -p1 <patches/sparse-block.diff
18 ./configure (optional if already run)
21 based-on: 2f84a6bd73e1a6e2081b1bcb965049ef83b2ce45
22 diff --git a/fileio.c b/fileio.c
26 #define ALIGNED_LENGTH(len) ((((len) - 1) | (ALIGN_BOUNDARY-1)) + 1)
28 extern int sparse_files;
29 +extern int sparse_files_block_size;
31 OFF_T preallocated_len = 0;
33 @@ -147,7 +148,7 @@ int write_file(int f, int use_seek, OFF_T offset, const char *buf, int len)
36 if (sparse_files > 0) {
37 - int len1 = MIN(len, SPARSE_WRITE_SIZE);
38 + int len1 = MIN(len, sparse_files_block_size);
39 r1 = write_sparse(f, use_seek, offset, buf, len1);
42 diff --git a/options.c b/options.c
45 @@ -79,6 +79,7 @@ int remove_source_files = 0;
46 int one_file_system = 0;
47 int protocol_version = PROTOCOL_VERSION;
49 +long sparse_files_block_size = SPARSE_WRITE_SIZE;
50 int preallocate_files = 0;
51 int do_compression = 0;
52 int def_compress_level = NOT_SPECIFIED;
53 @@ -732,6 +733,7 @@ void usage(enum logcode F)
54 rprintf(F," --fake-super store/recover privileged attrs using xattrs\n");
56 rprintf(F," -S, --sparse turn sequences of nulls into sparse blocks\n");
57 + rprintf(F," --sparse-block=SIZE set the block size used to handle sparse files\n");
58 #ifdef SUPPORT_PREALLOCATION
59 rprintf(F," --preallocate allocate dest files before writing them\n");
61 @@ -950,6 +952,7 @@ static struct poptOption long_options[] = {
62 {"sparse", 'S', POPT_ARG_VAL, &sparse_files, 1, 0, 0 },
63 {"no-sparse", 0, POPT_ARG_VAL, &sparse_files, 0, 0, 0 },
64 {"no-S", 0, POPT_ARG_VAL, &sparse_files, 0, 0, 0 },
65 + {"sparse-block", 0, POPT_ARG_LONG, &sparse_files_block_size, 0, 0, 0 },
66 {"preallocate", 0, POPT_ARG_NONE, &preallocate_files, 0, 0, 0},
67 {"inplace", 0, POPT_ARG_VAL, &inplace, 1, 0, 0 },
68 {"no-inplace", 0, POPT_ARG_VAL, &inplace, 0, 0, 0 },
69 @@ -2719,6 +2722,12 @@ void server_options(char **args, int *argc_p)
73 + if (sparse_files_block_size) {
74 + if (asprintf(&arg, "--sparse-block=%lu", sparse_files_block_size) < 0)
80 if (asprintf(&arg, "--timeout=%d", io_timeout) < 0)
82 diff --git a/rsync.yo b/rsync.yo
85 @@ -385,6 +385,7 @@ to the detailed description below for a complete description. verb(
86 --super receiver attempts super-user activities
87 --fake-super store/recover privileged attrs using xattrs
88 -S, --sparse turn sequences of nulls into sparse blocks
89 + --sparse-block=SIZE set block size used to handle sparse files
90 --preallocate allocate dest files before writing
91 --write-devices write to devices as files (implies --inplace)
92 -n, --dry-run perform a trial run with no changes made
93 @@ -1347,6 +1348,15 @@ If combined with bf(--sparse), the file will only have sparse blocks (as
94 opposed to allocated sequences of null bytes) if the kernel version and
95 filesystem type support creating holes in the allocated data.
97 +dit(bf(--sparse-block=SIZE)) Change the block size used to handle sparse files
98 +to SIZE bytes. This option only has an effect if the bf(--sparse) (bf(-S))
99 +option was also specified. The default block size used by rsync to detect a
100 +file hole is 1024 bytes; when the receiver writes data to the destination file
101 +and option bf(--sparse) is used, rsync checks every 1024-bytes chunk to detect
102 +if they are actually filled with data or not. With certain filesystems,
103 +optimized to receive data streams for example, enlarging this block size can
104 +strongly increase performance. The option can be used to tune this block size.
106 dit(bf(-n, --dry-run)) This makes rsync perform a trial run that doesn't
107 make any changes (and produces mostly the same output as a real run). It
108 is most commonly used in combination with the bf(-v, --verbose) and/or