Add support for preadv2/pwritev2

This is coming on Linux, and supports a high priority flag that will
use polling on the completion side.

Basic support is there, we'll need to add wrappers on x86_64 linux
to use them before it shows up in the glibc devel headers.

Signed-off-by: Jens Axboe <axboe@fb.com>

diff --git a/HOWTO b/HOWTO
index 5e765d4ca254fa945ee82a6009f7677b806a4aae..c37a9e09ac53d69d94fa925a5486b8fc8387eb02 100644 (file)
--- a/HOWTO
+++ b/HOWTO
@@ -1784,6 +1784,9 @@ that defines them is selected.
                enabled when polling for a minimum of 0 events (eg when
                iodepth_batch_complete=0).
 
+[psyncv2] hipri                Set RWF_HIPRI on IO, indicating to the kernel that
+                       it's of higher priority than normal.
+
 [cpu] cpuload=int Attempt to use the specified percentage of CPU cycles.
 
 [cpu] cpuchunks=int Split the load into cycles of the given time. In

diff --git a/configure b/configure
index cbd4d306abb1ed148c046e729548c87558720b65..6e2488ceff0e2a945ca3c996c92937ea82805752 100755 (executable)
--- a/configure
+++ b/configure
@@ -1240,6 +1240,22 @@ if compile_prog "" "" "pwritev"; then
 fi
 echo "pwritev/preadv                $pwritev"
 
+##########################################
+# Check whether we have pwritev2/preadv2
+pwritev2="no"
+cat > $TMPC << EOF
+#include <stdio.h>
+#include <sys/uio.h>
+int main(int argc, char **argv)
+{
+  return pwritev2(0, NULL, 1, 0, 0) + preadv2(0, NULL, 1, 0, 0);
+}
+EOF
+if compile_prog "" "" "pwritev2"; then
+  pwritev2="yes"
+fi
+echo "pwritev2/preadv2              $pwritev2"
+
 ##########################################
 # Check whether we have the required functions for ipv6
 ipv6="no"
@@ -1742,6 +1758,9 @@ fi
 if test "$pwritev" = "yes" ; then
   output_sym "CONFIG_PWRITEV"
 fi
+if test "$pwritev2" = "yes" ; then
+  output_sym "CONFIG_PWRITEV2"
+fi
 if test "$ipv6" = "yes" ; then
   output_sym "CONFIG_IPV6"
 fi

diff --git a/engines/sync.c b/engines/sync.c
index f5801fec9a9fd7a471e18d90fbb048d7a9fd03ec..0b0d1a756c3a8eafd0f74c7b6f990ae8331d90d2 100644 (file)
--- a/engines/sync.c
+++ b/engines/sync.c
@@ -13,6 +13,7 @@
 #include <assert.h>
 
 #include "../fio.h"
+#include "../optgroup.h"
 
 /*
  * Sync engine uses engine_data to store last offset
@@ -31,6 +32,28 @@ struct syncio_data {
        enum fio_ddir last_ddir;
 };
 
+#ifdef CONFIG_PWRITEV2
+struct psyncv2_options {
+       void *pad;
+       unsigned int hipri;
+};
+
+static struct fio_option options[] = {
+       {
+               .name   = "hipri",
+               .lname  = "RWF_HIPRI",
+               .type   = FIO_OPT_STR_SET,
+               .off1   = offsetof(struct psyncv2_options, hipri),
+               .help   = "Set RWF_HIPRI for pwritev2/preadv2",
+               .category = FIO_OPT_C_ENGINE,
+               .group  = FIO_OPT_G_INVALID,
+       },
+       {
+               .name   = NULL,
+       },
+};
+#endif
+
 static int fio_syncio_prep(struct thread_data *td, struct io_u *io_u)
 {
        struct fio_file *f = io_u->file;
@@ -98,6 +121,38 @@ static int fio_pvsyncio_queue(struct thread_data *td, struct io_u *io_u)
 }
 #endif
 
+#ifdef CONFIG_PWRITEV2
+static int fio_pvsyncio2_queue(struct thread_data *td, struct io_u *io_u)
+{
+       struct syncio_data *sd = td->io_ops->data;
+       struct psyncv2_options *o = td->eo;
+       struct iovec *iov = &sd->iovecs[0];
+       struct fio_file *f = io_u->file;
+       int ret, flags = 0;
+
+       fio_ro_check(td, io_u);
+
+       if (o->hipri)
+               flags |= RWF_HIPRI;
+
+       iov->iov_base = io_u->xfer_buf;
+       iov->iov_len = io_u->xfer_buflen;
+
+       if (io_u->ddir == DDIR_READ)
+               ret = preadv2(f->fd, iov, 1, io_u->offset, flags);
+       else if (io_u->ddir == DDIR_WRITE)
+               ret = pwritev2(f->fd, iov, 1, io_u->offset, flags);
+       else if (io_u->ddir == DDIR_TRIM) {
+               do_io_u_trim(td, io_u);
+               return FIO_Q_COMPLETED;
+       } else
+               ret = do_io_u_sync(td, io_u);
+
+       return fio_io_end(td, io_u, ret);
+}
+#endif
+
+
 static int fio_psyncio_queue(struct thread_data *td, struct io_u *io_u)
 {
        struct fio_file *f = io_u->file;
@@ -374,6 +429,22 @@ static struct ioengine_ops ioengine_pvrw = {
 };
 #endif
 
+#ifdef CONFIG_PWRITEV2
+static struct ioengine_ops ioengine_pvrw2 = {
+       .name           = "pvsync2",
+       .version        = FIO_IOOPS_VERSION,
+       .init           = fio_vsyncio_init,
+       .cleanup        = fio_vsyncio_cleanup,
+       .queue          = fio_pvsyncio2_queue,
+       .open_file      = generic_open_file,
+       .close_file     = generic_close_file,
+       .get_file_size  = generic_get_file_size,
+       .flags          = FIO_SYNCIO,
+       .options        = options,
+       .option_struct_size     = sizeof(struct psyncv2_options),
+};
+#endif
+
 static void fio_init fio_syncio_register(void)
 {
        register_ioengine(&ioengine_rw);

diff --git a/fio.1 b/fio.1
index 690c8f464e4e592d5a2be73edfa881313fc55015..f98802aadb9d209a0e273e352f4dfcc79f564726 100644 (file)
--- a/fio.1
+++ b/fio.1
@@ -591,6 +591,9 @@ coalescing adjacent IOs into a single submission.
 .B pvsync
 Basic \fBpreadv\fR\|(2) or \fBpwritev\fR\|(2) I/O.
 .TP
+.B pvsync2
+Basic \fBpreadv2\fR\|(2) or \fBpwritev2\fR\|(2) I/O.
+.TP
 .B libaio
 Linux native asynchronous I/O. This ioengine defines engine specific options.
 .TP
@@ -1647,6 +1650,10 @@ from user-space to reap events. The reaping mode is only
 enabled when polling for a minimum of 0 events (eg when
 iodepth_batch_complete=0).
 .TP
+.BI (psyncv2)hipri
+Set RWF_HIPRI on IO, indicating to the kernel that it's of
+higher priority than normal.
+.TP
 .BI (net,netsplice)hostname \fR=\fPstr
 The host name or IP address to use for TCP or UDP based IO.
 If the job is a TCP listener or UDP reader, the hostname is not

diff --git a/options.c b/options.c
index 390208729f0db12955cc3a19f72f1d04d9a40b1f..ac2da71f514ea22fcfe7b514b8dd0f6e36aa7f3c 100644 (file)
--- a/options.c
+++ b/options.c
@@ -1240,6 +1240,11 @@ struct fio_option fio_options[FIO_MAX_OPTS] = {
                            .help = "Use preadv/pwritev",
                          },
 #endif
+#ifdef CONFIG_PWRITEV
+                         { .ival = "pvsync2",
+                           .help = "Use preadv2/pwritev2",
+                         },
+#endif
 #ifdef CONFIG_LIBAIO
                          { .ival = "libaio",
                            .help = "Linux native asynchronous IO",