Re: [PATCH] vfs: Add a trace point in the mark_inode_dirty function

["Kok, Auke" <auke-jan.h.kok@intel.com>]

Ingo Molnar wrote:
> * Jeff Garzik <jeff@garzik.org> wrote:
> 
>> On 11/11/2009 01:34 AM, Arjan van de Ven wrote:
>>> Wu Fengguang<fengguang.wu@intel.com>  wrote:
>>>> Maybe this is enough for POWERTOP, however for general use, the dirty
>>>> type(data/metadata) and inode number may be valuable to some users?
>>> what can a user do with an inode number????
>> Inode numbers have always been visible to userspace...  IIRC, tar(1) 
>> uses the st_ino member of struct stat to detect hard links in certain 
>> cases.  ls(1) displays inode numbers with -i, find(1) looks for them 
>> with -inum, ...
> 
> Without an inode->vfs-name lookup/matching service it's of limited 
> utility though to developers and users. So inode numbers are fine (as 
> nicely unique physical identifiers)- as long as corresponding vfs name 
> string is available too.

agreed, this is the second problem I've stumbled upon (readahead was the first)
that could use something like this.

you can't reliably use inode numbers unless you actually know all of them at the
time that the tracepoint is generated: they could be deleted, modified etc. An
inode number is nice, but the filename and blockdev info would be superior for
both problems we're trying to solve.

Auke