[RFC PATCH] error out if ENOSPC during file layout

[RFC PATCH] error out if ENOSPC during file layout

[Kushal Kumaran]

Hi,

When I run fio with --create_only=1 and it runs out of space, it still
exits with status 0.  I could not figure out if this was intentional
(except for the fill_device case, where this is obviously the expected
behavior).

$ cat ~/fio 
[global]
ioengine=posixaio
rw=readwrite
size=2g
directory=${HOME}/mounts/testdisk
thread=1

[trivial-readwrite-1g]
$ df -h .
Filesystem      Size  Used Avail Use% Mounted on
/dev/loop6      976M  2.6M  907M   1% /home/kushal/mounts/testdisk
$ fio ~/fio --create_only=1
trivial-readwrite-1g: (g=0): rw=rw, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=pos
ixaio, iodepth=1                                                                                     
fio-3.12
Starting 1 thread
trivial-readwrite-1g: Laying out IO file (1 file / 2048MiB)
fio: ENOSPC on laying out file, stopping


Run status group 0 (all jobs):

Disk stats (read/write):
  loop6: ios=0/0, merge=0/0, ticks=0/0, in_queue=0, util=0.00%
$ echo $?
0

A trivial patch for this gives me the behavior I expect.

---
 filesetup.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/filesetup.c b/filesetup.c
index f4360a6f..42c5f630 100644
--- a/filesetup.c
+++ b/filesetup.c
@@ -231,13 +231,12 @@ static int extend_file(struct thread_data *td, struct fio_file *f)
 						break;
 					log_info("fio: ENOSPC on laying out "
 						 "file, stopping\n");
-					break;
 				}
 				td_verror(td, errno, "write");
 			} else
 				td_verror(td, EIO, "write");
 
-			break;
+			goto err;
 		}
 	}
 
-- 
2.20.1

Re: [RFC PATCH] error out if ENOSPC during file layout

[Sitsofe Wheeler]

Hi,

This does look like better behaviour to me.

Jens, can we pick this one up?

On Fri, 20 Nov 2020 at 05:09, Kushal Kumaran <kushal@locationd.net> wrote:
>
> Hi,
>
> When I run fio with --create_only=1 and it runs out of space, it still
> exits with status 0.  I could not figure out if this was intentional
> (except for the fill_device case, where this is obviously the expected
> behavior).
>
> $ cat ~/fio
> [global]
> ioengine=posixaio
> rw=readwrite
> size=2g
> directory=${HOME}/mounts/testdisk
> thread=1
>
> [trivial-readwrite-1g]
> $ df -h .
> Filesystem      Size  Used Avail Use% Mounted on
> /dev/loop6      976M  2.6M  907M   1% /home/kushal/mounts/testdisk
> $ fio ~/fio --create_only=1
> trivial-readwrite-1g: (g=0): rw=rw, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=pos
> ixaio, iodepth=1
> fio-3.12
> Starting 1 thread
> trivial-readwrite-1g: Laying out IO file (1 file / 2048MiB)
> fio: ENOSPC on laying out file, stopping
>
>
> Run status group 0 (all jobs):
>
> Disk stats (read/write):
>   loop6: ios=0/0, merge=0/0, ticks=0/0, in_queue=0, util=0.00%
> $ echo $?
> 0
>
> A trivial patch for this gives me the behavior I expect.
>
> ---
>  filesetup.c | 3 +--
>  1 file changed, 1 insertion(+), 2 deletions(-)
>
> diff --git a/filesetup.c b/filesetup.c
> index f4360a6f..42c5f630 100644
> --- a/filesetup.c
> +++ b/filesetup.c
> @@ -231,13 +231,12 @@ static int extend_file(struct thread_data *td, struct fio_file *f)
>                                                 break;
>                                         log_info("fio: ENOSPC on laying out "
>                                                  "file, stopping\n");
> -                                       break;
>                                 }
>                                 td_verror(td, errno, "write");
>                         } else
>                                 td_verror(td, EIO, "write");
>
> -                       break;
> +                       goto err;
>                 }
>         }
>
> --
> 2.20.1



-- 
Sitsofe | http://sucs.org/~sits/

Re: [RFC PATCH] error out if ENOSPC during file layout

[Jens Axboe]

On 11/19/20 10:00 PM, Kushal Kumaran wrote:
> Hi,
> 
> When I run fio with --create_only=1 and it runs out of space, it still
> exits with status 0.  I could not figure out if this was intentional
> (except for the fill_device case, where this is obviously the expected
> behavior).
> 
> $ cat ~/fio 
> [global]
> ioengine=posixaio
> rw=readwrite
> size=2g
> directory=${HOME}/mounts/testdisk
> thread=1
> 
> [trivial-readwrite-1g]
> $ df -h .
> Filesystem      Size  Used Avail Use% Mounted on
> /dev/loop6      976M  2.6M  907M   1% /home/kushal/mounts/testdisk
> $ fio ~/fio --create_only=1
> trivial-readwrite-1g: (g=0): rw=rw, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=pos
> ixaio, iodepth=1                                                                                     
> fio-3.12
> Starting 1 thread
> trivial-readwrite-1g: Laying out IO file (1 file / 2048MiB)
> fio: ENOSPC on laying out file, stopping
> 
> 
> Run status group 0 (all jobs):
> 
> Disk stats (read/write):
>   loop6: ios=0/0, merge=0/0, ticks=0/0, in_queue=0, util=0.00%
> $ echo $?
> 0
> 
> A trivial patch for this gives me the behavior I expect.

That seems reasonable. Applied, thanks.

-- 
Jens Axboe