Add ability to invoke fallocate() FALLOC_FL_KEEP_SIZE.

Linux offers fallocate() and the FALLOC_FL_KEEP_SIZE option as
an alternative to posix_fallocate(). When FALLOC_FL_KEEP_SIZE is
specified for an falloc request going beyond the end of the file,
the requested blocks get preallocated without changing the apparent
size of the file. This is is a commonly recommended use of fallocate()
for workloads performing append writes.

This patch modifies the fallocate option from a boolean option
to a string option accepting none/posix/keep/0/1. 'keep' is only
made available on systems where FALLOC_FL_KEEP_SIZE is available
(i.e., Linux at this time). If specified, fallocate() is used
with FALLOC_FL_KEEP_SIZE set. 'none' disables pre-allocation while
'posix' uses posix_fallocate(). The default behavior remains unchaned,
i.e., invoking posix_fallocate. The settings '0'/'1' are there to
provide backward compatibility for users who had explicitly set the
boolean option.

Signed-off-by: Jens Axboe <jaxboe@fusionio.com>

diff --git a/HOWTO b/HOWTO
index 69b8cc6766850b6ceab07a0c15ea7e872142838e..ee899b8fdb6c402814086dde41f986d152c456c9 100644 (file)
--- a/HOWTO
+++ b/HOWTO
@@ -354,11 +354,19 @@ use_os_rand=bool Fio can either use the random generator supplied by the OS
                internal generator, which is often of better quality and
                faster.
 
-fallocate=bool By default, fio will use fallocate() to advise the system
-               of the size of the file we are going to write. This can be
-               turned off with fallocate=0. May not be available on all
-               supported platforms.  If using ZFS on Solaris this must be
-               set to 0 because ZFS doesn't support it.
+fallocate=str  Whether pre-allocation is performed when laying down files.
+               Accepted values are:
+
+                       none            Do not pre-allocate space
+                       posix           Pre-allocate via posix_fallocate()
+                       keep            Pre-allocate via fallocate() with
+                                       FALLOC_FL_KEEP_SIZE set
+                       0               Backward-compatible alias for 'none'
+                       1               Backward-compatible alias for 'posix'
+
+               May not be available on all supported platforms. 'keep' is only
+               available on Linux.If using ZFS on Solaris this must be set to
+               'none' because ZFS doesn't support it. Default: 'posix'.
 
 fadvise_hint=bool By default, fio will use fadvise() to advise the kernel
                on what IO patterns it is likely to issue. Sometimes you

diff --git a/file.h b/file.h
index 04c0d4536b4f16aa91ca58609de80c1f973228c7..b3ff05148458e1a1d0d5a70e0c0fd874fa892b18 100644 (file)
--- a/file.h
+++ b/file.h
@@ -42,6 +42,16 @@ enum {
        FIO_FSERVICE_SEQ        = 3,
 };
 
+/*
+ * No pre-allocation when laying down files, or call posix_fallocate(), or
+ * call fallocate() with FALLOC_FL_KEEP_SIZE set.
+ */
+enum fio_fallocate_mode {
+       FIO_FALLOCATE_NONE      = 1,
+       FIO_FALLOCATE_POSIX     = 2,
+       FIO_FALLOCATE_KEEP_SIZE = 3,
+};
+
 /*
  * Each thread_data structure has a number of files associated with it,
  * this structure holds state information for a single file.

diff --git a/filesetup.c b/filesetup.c
index 799202f0da5bbccf9b7aca0e90fc7f7ed5293004..6d8aa7a7087bdc67c8e72d0f5a8a4aa9fe22b8ef 100644 (file)
--- a/filesetup.c
+++ b/filesetup.c
@@ -13,6 +13,10 @@
 #include "filehash.h"
 #include "os/os.h"
 
+#ifdef FIO_HAVE_LINUX_FALLOCATE
+#include <linux/falloc.h>
+#endif
+
 static int root_warn;
 
 static inline void clear_error(struct thread_data *td)
@@ -67,17 +71,41 @@ static int extend_file(struct thread_data *td, struct fio_file *f)
        }
 
 #ifdef FIO_HAVE_FALLOCATE
-       if (td->o.fallocate && !td->o.fill_device) {
-               dprint(FD_FILE, "fallocate file %s size %llu\n", f->file_name,
-                                                       f->real_file_size);
-
-               r = posix_fallocate(f->fd, 0, f->real_file_size);
-               if (r > 0) {
-                       log_err("fio: posix_fallocate fails: %s\n",
-                                       strerror(r));
+       if (!td->o.fill_device) {
+               switch (td->o.fallocate_mode) {
+               case FIO_FALLOCATE_NONE:
+                       break;
+               case FIO_FALLOCATE_POSIX:
+                       dprint(FD_FILE, "posix_fallocate file %s size %llu\n",
+                                f->file_name, f->real_file_size);
+
+                       r = posix_fallocate(f->fd, 0, f->real_file_size);
+                       if (r > 0) {
+                               log_err("fio: posix_fallocate fails: %s\n",
+                                               strerror(r));
+                       }
+                       break;
+#ifdef FIO_HAVE_LINUX_FALLOCATE
+               case FIO_FALLOCATE_KEEP_SIZE:
+                       dprint(FD_FILE,
+                               "fallocate(FALLOC_FL_KEEP_SIZE) "
+                               "file %s size %llu\n",
+                               f->file_name, f->real_file_size);
+
+                       r = fallocate(f->fd, FALLOC_FL_KEEP_SIZE, 0,
+                                       f->real_file_size);
+                       if (r != 0) {
+                               td_verror(td, errno, "fallocate");
+                       }
+                       break;
+#endif /* FIO_HAVE_LINUX_FALLOCATE */
+               default:
+                       log_err("fio: unknown fallocate mode: %d\n",
+                               td->o.fallocate_mode);
+                       assert(0);
                }
        }
-#endif
+#endif /* FIO_HAVE_FALLOCATE */
 
        if (!new_layout)
                goto done;

diff --git a/fio.1 b/fio.1
index 0ced604de4f82b0fadb7345b34e61735d46a0b01..ad5040b37372e93415b7783db9924934211213cf 100644 (file)
--- a/fio.1
+++ b/fio.1
@@ -220,10 +220,32 @@ offsets, or it can use it's own internal generator (based on Tausworthe).
 Default is to use the internal generator, which is often of better quality and
 faster. Default: false.
 .TP
-.BI fallocate \fR=\fPbool
-By default, fio will use fallocate() to advise the system of the size of the
-file we are going to write. This can be turned off with fallocate=0. May not
-be available on all supported platforms.
+.BI fallocate \fR=\fPstr
+Whether pre-allocation is performed when laying down files. Accepted values
+are:
+.RS
+.RS
+.TP
+.B none
+Do not pre-allocate space.
+.TP
+.B posix
+Pre-allocate via posix_fallocate().
+.TP
+.B keep
+Pre-allocate via fallocate() with FALLOC_FL_KEEP_SIZE set.
+.TP
+.B 0
+Backward-compatible alias for 'none'.
+.TP
+.B 1
+Backward-compatible alias for 'posix'.
+.RE
+.P
+May not be available on all supported platforms. 'keep' is only
+available on Linux. If using ZFS on Solaris this must be set to 'none'
+because ZFS doesn't support it. Default: 'posix'.
+.RE
 .TP
 .BI fadvise_hint \fR=\fPbool
 Disable use of \fIposix_fadvise\fR\|(2) to advise the kernel what I/O patterns

diff --git a/fio.h b/fio.h
index 6ad186f0e6a87cc4ff5ec20dda35c598df0aaf8c..16866ddff95737805abc5610be8e471bc805ca39 100644 (file)
--- a/fio.h
+++ b/fio.h
@@ -248,7 +248,7 @@ struct thread_options {
        unsigned int file_service_type;
        unsigned int group_reporting;
        unsigned int fadvise_hint;
-       unsigned int fallocate;
+       enum fio_fallocate_mode fallocate_mode;
        unsigned int zero_buffers;
        unsigned int refill_buffers;
        unsigned int time_based;

diff --git a/options.c b/options.c
index a9b05349a808d09b0596202c3d015eeca2277229..bd7dc99fddeffa7c74c3d51fa945fa62fe1763e1 100644 (file)
--- a/options.c
+++ b/options.c
@@ -1178,12 +1178,37 @@ static struct fio_option options[FIO_MAX_OPTS] = {
 #ifdef FIO_HAVE_FALLOCATE
        {
                .name   = "fallocate",
-               .type   = FIO_OPT_BOOL,
-               .off1   = td_var_offset(fallocate),
-               .help   = "Use fallocate() when laying out files",
-               .def    = "1",
-       },
+               .type   = FIO_OPT_STR,
+               .off1   = td_var_offset(fallocate_mode),
+               .help   = "Whether pre-allocation is performed when laying out files",
+               .def    = "posix",
+               .posval = {
+                         { .ival = "none",
+                           .oval = FIO_FALLOCATE_NONE,
+                           .help = "Do not pre-allocate space",
+                         },
+                         { .ival = "posix",
+                           .oval = FIO_FALLOCATE_POSIX,
+                           .help = "Use posix_fallocate()",
+                         },
+#ifdef FIO_HAVE_LINUX_FALLOCATE
+                         { .ival = "keep",
+                           .oval = FIO_FALLOCATE_KEEP_SIZE,
+                           .help = "Use fallocate(..., FALLOC_FL_KEEP_SIZE, ...)",
+                         },
 #endif
+                         /* Compatibility with former boolean values */
+                         { .ival = "0",
+                           .oval = FIO_FALLOCATE_NONE,
+                           .help = "Alias for 'none'",
+                         },
+                         { .ival = "1",
+                           .oval = FIO_FALLOCATE_POSIX,
+                           .help = "Alias for 'posix'",
+                         },
+               },
+       },
+#endif /* FIO_HAVE_FALLOCATE */
        {
                .name   = "fadvise_hint",
                .type   = FIO_OPT_BOOL,

diff --git a/os/os-linux.h b/os/os-linux.h
index 70c993b5de61d449051cfff989482a728cdfd2ad..024ef89ae23d1d4548b6c4a7982de9db1b6913b2 100644 (file)
--- a/os/os-linux.h
+++ b/os/os-linux.h
@@ -32,6 +32,7 @@
 #define FIO_HAVE_BLKTRACE
 #define FIO_HAVE_STRSEP
 #define FIO_HAVE_FALLOCATE
+#define FIO_HAVE_LINUX_FALLOCATE
 #define FIO_HAVE_POSIXAIO_FSYNC
 #define FIO_HAVE_PSHARED_MUTEX
 #define FIO_HAVE_CL_SIZE