Subject: Re: msdosfs performance problem
To: None <>
From: Christos Zoulas <>
List: tech-kern
Date: 11/28/2005 00:24:06
In article <>,
Matthew Orgass  <> wrote:
>On 2005-11-27 wrote:
>> (Matthew Orgass) writes:
>> >  msdosfs does all writes fully synchronously (why, I can't say).
>> As I understand it meta data writes are synchronous but
>> can be asynchronous if you mount -o async. Data writes
>> are done by UBC and are asynchronous.
>  You are right, it isn't fully synchronous. I should have tested it more
>(checking directly now does show serious performance loss with -o sync).
>However it doesn't seem to behave quite like normal ffs in that all data
>will be flushed before the writing command completes, and dd seems to show
>accurate times early in the transfer.

And performance degrades dramatically as the file size increases.