User: Password:
|
|
Subscribe / Log in / New account

Fixing the unfixable autofs ABI

Fixing the unfixable autofs ABI

Posted May 1, 2012 0:36 UTC (Tue) by hamjudo (guest, #363)
In reply to: Fixing the unfixable autofs ABI by ewan
Parent article: Fixing the unfixable autofs ABI

This is a two part kludge. Part one is adding the "unusual" behavior to pipes. Part two, is making it so that the kernel side of the autofs code will set the flag in the right situations.

The history of operating systems is filled with examples of bad design decisions that weren't caught in time, and thus got stuck in the interface definition.


(Log in to post comments)

Fixing the unfixable autofs ABI

Posted May 1, 2012 0:55 UTC (Tue) by corbet (editor, #1) [Link]

Almost. The kernel side always turns on packetized behavior so that things work as expected with all known user-space binaries.

Fixing the unfixable autofs ABI

Posted May 1, 2012 15:08 UTC (Tue) by ewan (subscriber, #5533) [Link]

I'm sure I'm being dense, but when the article says "Linus made an open with O_DIRECT turn on the packetized behavior" surely that rather implies that it's off unless you use O_DIRECT?

Fixing the unfixable autofs ABI

Posted May 1, 2012 15:38 UTC (Tue) by hamjudo (guest, #363) [Link]

It is off unless, something turns on O_DIRECT. The kernel side of the autofs interface turns on O_DIRECT. As far as I know, it isn't turned on anywhere else in the kernel (yet).

Nothing, but common sense, prevents users from writing programs that use this feature.

Fixing the unfixable autofs ABI

Posted May 1, 2012 15:41 UTC (Tue) by juliank (subscriber, #45896) [Link]

As far as I understand it, if you create such a a pipe yourself, then you use O_DIRECT; and for autofs, the kernel creates such a pipe itself and you thus automatically get that.


Copyright © 2017, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds