Skip to contentSkip to conversation

procfs feature or bug?

DISCUSSION Sandip Bhattacharya 2 participants 2 messages Sep 19, 2001

lore
Share & tools
lore
#0 lore raw

[Do cc me replies. Thanks.]

Hi!

I was trying out the multipage procfs entries, when i found out that i was having a problem.

I am trying to use the "hack" by Paul Russell to allow mangling of filepos using ``*start'' entries as my personal page offset. Now, for multipage entries, whatever i set as *start should be coming back to me as offset when the next page is called.

But in fs/proc/generic.c in proc_file_read() at the end we have ( after the first page has been "copied_to_user")

*ppos += start < page ? (long) start : n;

But this _adds_ the contents of start to the offset, in the case where I am supplying the offset in ``start''. Shouldn't this just be _replacing_ ? 'Cause in this case the offsets get cumulatively added, causing an oops at the end for me :(

Or am I missing something big???

- Sandip

--
-----------------------------------------------------------
Sandip Bhattacharya
Office: [email protected] Home: [email protected]
Mindframe Software
-----------------------------------------------------------

#1 lore raw
Quoted 11 lines · Sandip Bhattacharya

I was trying out the multipage procfs entries, when i found out that i was having a problem.

I am trying to use the "hack" by Paul Russell to allow mangling of filepos using ``*start'' entries as my personal page offset. Now, for multipage entries, whatever i set as *start should be coming back to me as offset when the next page is called.

But in fs/proc/generic.c in proc_file_read() at the end we have ( after the first page has been "copied_to_user")

Quoted 8 lines

*ppos += start < page ? (long) start : n;

But this _adds_ the contents of start to the offset, in the case where I am supplying the offset in ``start''. Shouldn't this just be _replacing_ ? 'Cause in this case the offsets get cumulatively added, causing an oops at the end for me :(

Or am I missing something big???

Per the comment:

Quoted 5 lines
		/* This is a hack to allow mangling of file pos independent
(quoted diff) +0 -0
 		 * of actual bytes read.  Simply place the data at page,
 		 * return the bytes, and set `start' to the desired offset
 		 * as an unsigned int. - [email protected]
		*/

*ppos represents the offset in the *file*, not the buffer, so cumulative is correct. So when you do your read, you need to interpret it thus. I've been using it in one case as a record number, where the records are variable length. I set *start to 1, and treat is as an index into a (virtual) array of records.

Signature · 1 line
/Jonathan Lundell.