Groups > comp.os.linux.development.apps > #488

Re: file I/O: writ to disk in parallel?

From	Lusotec <nomail@nomail.not>
Newsgroups	comp.os.linux.development.apps
Subject	Re: file I/O: writ to disk in parallel?
Followup-To	comp.os.linux.development.apps
Date	2012-06-21 13:33 +0100
Organization	A noiseless patient Spider
Message-ID	<jrv4b7$f6n$1@dont-email.me> (permalink)
References	<c7cb8f1d-73ec-45ed-9bbe-b06a6e5c140c@googlegroups.com>

Followups directed to: comp.os.linux.development.apps

Show all headers | View raw

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

lmike wrote:
> Say an application code calls fwrite/write etc. to write a very long
> string "aaa ... 123 ... ccc" to a disk file, is it ever possible that the
> app/box crashes in the middle of the write operation, then it leaves a
> garbage like "aaa ... @#$@#$ ... ccc" or "@#$%@#$ ... 123 ... ccc" in the
> disk file?
> 
> Or, in other words, could the disk write head(s) (1) write parts of that
> string in parallel to the disk, OR (2) write the bytes not in order, e.g.
> first write "ccc", then "aaa" ... ? Any hard disks support that? Thanks!

fwrite/write gives no guarantees on write order or if it is actually written 
to the device. This is true even if there are no crashes or power issues. If 
you needs *some* guarantees then look in to fsync.

Regards.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)

iF4EAREIAAYFAk/jFKQACgkQGQjO2ccW76qgzgD/S08n8W2Yct+sqajgN8o82sxt
1m7RAVPnCgfGlMfzfQEA/05N6X5sJfymo68o1yvNVlPn0SNs14uZ0cJy4zi3ffgF
=KyMW
-----END PGP SIGNATURE-----

Back to comp.os.linux.development.apps | Previous | Next — Previous in thread | Next in thread | Find similar

Thread

file I/O: writ to disk in parallel? lmike <lmike3000@gmail.com> - 2012-06-20 20:35 -0700
  Re: file I/O: writ to disk in parallel? Lusotec <nomail@nomail.not> - 2012-06-21 13:33 +0100
  Re: file I/O: writ to disk in parallel? Rainer Weikusat <rweikusat@mssgmbh.com> - 2012-06-21 14:17 +0100

csiph-web