zipf/pareto: use min of file size and io size for init

[fio.git] / filesetup.c

diff --git a/filesetup.c b/filesetup.c
index ac69c65ea8f44956debb01bca32bef41c2675fb2..41809093ffd92b8eac8f5b2e5e89f16facfab12a 100644 (file)
--- a/filesetup.c
+++ b/filesetup.c
@@ -12,6 +12,8 @@
 #include "smalloc.h"
 #include "filehash.h"
 #include "os/os.h"
+#include "hash.h"
+#include "lib/axmap.h"
 
 #ifdef FIO_HAVE_LINUX_FALLOCATE
 #include <linux/falloc.h>
@@ -432,6 +434,12 @@ int generic_close_file(struct thread_data fio_unused *td, struct fio_file *f)
                ret = errno;
 
        f->fd = -1;
+
+       if (f->shadow_fd != -1) {
+               close(f->shadow_fd);
+               f->shadow_fd = -1;
+       }
+
        return ret;
 }
 
@@ -460,6 +468,24 @@ int file_lookup_open(struct fio_file *f, int flags)
        return from_hash;
 }
 
+static int file_close_shadow_fds(struct thread_data *td)
+{
+       struct fio_file *f;
+       int num_closed = 0;
+       unsigned int i;
+
+       for_each_file(td, f, i) {
+               if (f->shadow_fd == -1)
+                       continue;
+
+               close(f->shadow_fd);
+               f->shadow_fd = -1;
+               num_closed++;
+       }
+
+       return num_closed;
+}
+
 int generic_open_file(struct thread_data *td, struct fio_file *f)
 {
        int is_std = 0;
@@ -534,6 +560,8 @@ open_again:
                        flags &= ~FIO_O_NOATIME;
                        goto open_again;
                }
+               if (__e == EMFILE && file_close_shadow_fds(td))
+                       goto open_again;
 
                snprintf(buf, sizeof(buf) - 1, "open(%s)", f->file_name);
 
@@ -550,9 +578,22 @@ open_again:
                        int fio_unused ret;
 
                        /*
-                        * OK to ignore, we haven't done anything with it
+                        * Stash away descriptor for later close. This is to
+                        * work-around a "feature" on Linux, where a close of
+                        * an fd that has been opened for write will trigger
+                        * udev to call blkid to check partitions, fs id, etc.
+                        * That polutes the device cache, which can slow down
+                        * unbuffered accesses.
                         */
-                       ret = generic_close_file(td, f);
+                       if (f->shadow_fd == -1)
+                               f->shadow_fd = f->fd;
+                       else {
+                               /*
+                                * OK to ignore, we haven't done anything
+                                * with it
+                                */
+                               ret = generic_close_file(td, f);
+                       }
                        goto open_again;
                }
        }
@@ -864,17 +905,23 @@ int pre_read_files(struct thread_data *td)
 
 static int __init_rand_distribution(struct thread_data *td, struct fio_file *f)
 {
-       unsigned int range_size;
+       unsigned int range_size, seed;
        unsigned long nranges;
+       uint64_t file_size;
 
        range_size = min(td->o.min_bs[DDIR_READ], td->o.min_bs[DDIR_WRITE]);
+       file_size = min(f->real_file_size, f->io_size);
+
+       nranges = (file_size + range_size - 1) / range_size;
 
-       nranges = (f->real_file_size + range_size - 1) / range_size;
+       seed = jhash(f->file_name, strlen(f->file_name), 0) * td->thread_number;
+       if (!td->o.rand_repeatable)
+               seed = td->rand_seeds[4];
 
        if (td->o.random_distribution == FIO_RAND_DIST_ZIPF)
-               zipf_init(&f->zipf, nranges, td->o.zipf_theta);
+               zipf_init(&f->zipf, nranges, td->o.zipf_theta, seed);
        else
-               pareto_init(&f->zipf, nranges, td->o.pareto_h);
+               pareto_init(&f->zipf, nranges, td->o.pareto_h, seed);
 
        return 1;
 }
@@ -899,28 +946,31 @@ static int init_rand_distribution(struct thread_data *td)
 
 int init_random_map(struct thread_data *td)
 {
-       unsigned long long blocks, num_maps;
+       unsigned long long blocks;
        struct fio_file *f;
        unsigned int i;
 
        if (init_rand_distribution(td))
                return 0;
-       if (td->o.norandommap || !td_random(td))
+       if (!td_random(td))
                return 0;
 
        for_each_file(td, f, i) {
                blocks = (f->real_file_size + td->o.rw_min_bs - 1) /
                                (unsigned long long) td->o.rw_min_bs;
-               num_maps = (blocks + BLOCKS_PER_MAP - 1) /
-                               (unsigned long long) BLOCKS_PER_MAP;
-               if (num_maps == (unsigned long) num_maps) {
-                       f->file_map = smalloc(num_maps * sizeof(unsigned long));
-                       if (f->file_map) {
-                               f->num_maps = num_maps;
+               if (td->o.random_generator == FIO_RAND_GEN_LFSR) {
+                       unsigned long seed;
+
+                       seed = td->rand_seeds[FIO_RAND_BLOCK_OFF];
+                       
+                       if (!lfsr_init(&f->lfsr, blocks, seed))
                                continue;
-                       }
-               } else
-                       f->file_map = NULL;
+               } else if (!td->o.norandommap) {
+                       f->io_axmap = axmap_new(blocks);
+                       if (f->io_axmap)
+                               continue;
+               } else if (td->o.norandommap)
+                       continue;
 
                if (!td->o.softrandommap) {
                        log_err("fio: failed allocating random map. If running"
@@ -932,7 +982,6 @@ int init_random_map(struct thread_data *td)
 
                log_info("fio: file %s failed allocating random map. Running "
                         "job without.\n", f->file_name);
-               f->num_maps = 0;
        }
 
        return 0;
@@ -969,8 +1018,8 @@ void close_and_free_files(struct thread_data *td)
 
                sfree(f->file_name);
                f->file_name = NULL;
-               sfree(f->file_map);
-               f->file_map = NULL;
+               axmap_free(f->io_axmap);
+               f->io_axmap = NULL;
                sfree(f);
        }
 
@@ -1021,6 +1070,7 @@ int add_file(struct thread_data *td, const char *fname)
        }
 
        f->fd = -1;
+       f->shadow_fd = -1;
        fio_file_reset(f);
 
        if (td->files_size <= td->files_index) {