tech-kern archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]


Thank you David, the provided link was helpful.

I never really did understand the PNBUF_PUT calls but simply mimicked other 
filesystem implementations. I used tmpfs a lot in this regard. Since SAVESTART 
and SAVENAME was eliminated, it allowed me to discard all the explicit 
PNBUF_PUT calls since they were dependant on SAVESTART and SAVENAME flags.

Vnode locking, interlock, and usecount is still complicated to me but I think I 
have a handle on it now. The vnode man page does not explicit state the side 
effects of the various vnode functions with regard to vnode lock, interlock, 
and usecount. But I was able to muddle thru the source of vfs_vnode.c to 
discern the side effects that I cared about. 

I was stung by the fact that I used system structures in my on media inode 
representation. That has been corrected.

Thanks for your help.

On Jan 16, 2013, at 9:07 PM, David Holland wrote:

> On Wed, Jan 16, 2013 at 04:24:15PM -0600, Frank Zerangue wrote:
>> Migrating from 5.1.2 to 6.0.1 I have noticed many changes in the
>> virtual filesystem support interface. Can anyone point me to some
>> discussion or documentation of these changes?
> That is oriented towards people maintaining out-of-tree filesystems,
> but is supposed to at least mention all the incompatible changes.
> If you have further questions, ask away.
>> Can anyone give me some insight in regard to the elimination of
>> PNBUF_PUT in tmpfs?
> The handling of path buffers in namei and the namei results structure
> was cleaned up a good deal. The SAVESTART and SAVENAME flags were both
> eliminated entirely.
> -- 
> David A. Holland

Frank Zerangue

Home | Main Index | Thread Index | Old Index